]> git.lyx.org Git - lyx.git/blob - src/LaTeX.cpp
Fixup 572b06d6: reduce cache size for breakString
[lyx.git] / src / LaTeX.cpp
1 /**
2  * \file LaTeX.cpp
3  * This file is part of LyX, the document processor.
4  * Licence details can be found in the file COPYING.
5  *
6  * \author Alfredo Braunstein
7  * \author Lars Gullik Bjønnes
8  * \author Jean-Marc Lasgouttes
9  * \author Angus Leeming
10  * \author Dekel Tsur
11  * \author Jürgen Spitzmüller
12  *
13  * Full author contact details are available in file CREDITS.
14  */
15
16 #include <config.h>
17
18 #include "LaTeX.h"
19
20 #include "Buffer.h"
21 #include "BufferList.h"
22 #include "BufferParams.h"
23 #include "LyXRC.h"
24 #include "LyX.h"
25 #include "DepTable.h"
26 #include "Encoding.h"
27 #include "Language.h"
28
29 #include "support/debug.h"
30 #include "support/docstring.h"
31 #include "support/convert.h"
32 #include "support/FileName.h"
33 #include "support/filetools.h"
34 #include "support/gettext.h"
35 #include "support/lstrings.h"
36 #include "support/Systemcall.h"
37 #include "support/os.h"
38
39 #include <fstream>
40 #include <regex>
41 #include <stack>
42
43
44 using namespace std;
45 using namespace lyx::support;
46
47 namespace lyx {
48
49 namespace os = support::os;
50
51 // TODO: in no particular order
52 // - get rid of the call to
53 //   BufferList::updateIncludedTeXfiles, this should either
54 //   be done before calling LaTeX::funcs or in a completely
55 //   different way.
56 // - the makeindex style files should be taken care of with
57 //   the dependency mechanism.
58
59 namespace {
60
61 docstring runMessage(unsigned int count)
62 {
63         return bformat(_("Waiting for LaTeX run number %1$d"), count);
64 }
65
66 } // namespace
67
68 /*
69  * CLASS TEXERRORS
70  */
71
72 void TeXErrors::insertError(int line, docstring const & error_desc,
73                             docstring const & error_text,
74                             string const & child_name)
75 {
76         Error newerr(line, error_desc, error_text, child_name);
77         errors.push_back(newerr);
78 }
79
80
81 void TeXErrors::insertRef(int line, docstring const & error_desc,
82                             docstring const & error_text,
83                             string const & child_name)
84 {
85         Error newerr(line, error_desc, error_text, child_name);
86         undef_ref.push_back(newerr);
87 }
88
89
90 bool operator==(AuxInfo const & a, AuxInfo const & o)
91 {
92         return a.aux_file == o.aux_file
93                 && a.citations == o.citations
94                 && a.databases == o.databases
95                 && a.styles == o.styles;
96 }
97
98
99 bool operator!=(AuxInfo const & a, AuxInfo const & o)
100 {
101         return !(a == o);
102 }
103
104
105 /*
106  * CLASS LaTeX
107  */
108
109 LaTeX::LaTeX(string const & latex, OutputParams const & rp,
110              FileName const & f, string const & p, string const & lp, 
111              bool allow_cancellation, bool const clean_start)
112         : cmd(latex), file(f), path(p), lpath(lp), runparams(rp), biber(false),
113           allow_cancel(allow_cancellation)
114 {
115         num_errors = 0;
116         // lualatex can still produce a DVI with --output-format=dvi. However,
117         // we do not use that internally (we use the "dvilualatex" command) so
118         // it would only happen from a custom converter. Thus, it is better to
119         // guess that lualatex produces a PDF than to guess a DVI.
120         // FIXME we should base the extension on the output format, which we should
121         // get in a robust way, e.g. from the converter.
122         if (prefixIs(cmd, "pdf") || prefixIs(cmd, "lualatex") || prefixIs(cmd, "xelatex")) {
123                 depfile = FileName(file.absFileName() + ".dep-pdf");
124                 output_file =
125                         FileName(changeExtension(file.absFileName(), ".pdf"));
126         } else {
127                 depfile = FileName(file.absFileName() + ".dep");
128                 output_file =
129                         FileName(changeExtension(file.absFileName(), ".dvi"));
130         }
131         if (clean_start)
132                 removeAuxiliaryFiles();
133 }
134
135
136 void LaTeX::removeAuxiliaryFiles() const
137 {
138         LYXERR(Debug::OUTFILE, "Removing auxiliary files");
139         // Note that we do not always call this function when there is an error.
140         // For example, if there is an error but an output file is produced we
141         // still would like to output (export/view) the file.
142
143         // What files do we have to delete?
144
145         // This will at least make latex do all the runs
146         depfile.removeFile();
147
148         // but the reason for the error might be in a generated file...
149
150         // bibtex file
151         FileName const bbl(changeExtension(file.absFileName(), ".bbl"));
152         bbl.removeFile();
153
154         // biber file
155         FileName const bcf(changeExtension(file.absFileName(), ".bcf"));
156         bcf.removeFile();
157
158         // makeindex file
159         FileName const ind(changeExtension(file.absFileName(), ".ind"));
160         ind.removeFile();
161
162         // nomencl file
163         FileName const nls(changeExtension(file.absFileName(), ".nls"));
164         nls.removeFile();
165
166         // nomencl file (old version of the package)
167         FileName const gls(changeExtension(file.absFileName(), ".gls"));
168         gls.removeFile();
169
170         // endnotes file
171         FileName const ent(changeExtension(file.absFileName(), ".ent"));
172         ent.removeFile();
173
174         // Also remove the aux file
175         FileName const aux(changeExtension(file.absFileName(), ".aux"));
176         aux.removeFile();
177
178         // Also remove the .out file (e.g. hyperref bookmarks) (#9963)
179         FileName const out(changeExtension(file.absFileName(), ".out"));
180         out.removeFile();
181
182         // Remove the output file, which is often generated even if error
183         output_file.removeFile();
184 }
185
186
187 int LaTeX::run(TeXErrors & terr)
188         // We know that this function will only be run if the lyx buffer
189         // has been changed. We also know that a newly written .tex file
190         // is always different from the previous one because of the date
191         // in it. However it seems safe to run latex (at least) one time
192         // each time the .tex file changes.
193 {
194         int scanres = NO_ERRORS;
195         int bscanres = NO_ERRORS;
196         int iscanres = NO_ERRORS;
197         unsigned int count = 0; // number of times run
198         num_errors = 0; // just to make sure.
199         unsigned int const MAX_RUN = 6;
200         DepTable head; // empty head
201         bool rerun = false; // rerun requested
202
203         // The class LaTeX does not know the temp path.
204         theBufferList().updateIncludedTeXfiles(FileName::getcwd().absFileName(),
205                 runparams);
206
207         // 0
208         // first check if the file dependencies exist:
209         //     ->If it does exist
210         //             check if any of the files mentioned in it have
211         //             changed (done using a checksum).
212         //                 -> if changed:
213         //                        run latex once and
214         //                        remake the dependency file
215         //                 -> if not changed:
216         //                        just return there is nothing to do for us.
217         //     ->if it doesn't exist
218         //             make it and
219         //             run latex once (we need to run latex once anyway) and
220         //             remake the dependency file.
221         //
222
223         bool had_depfile = depfile.exists();
224         bool run_bibtex = false;
225         FileName const aux_file(changeExtension(file.absFileName(), ".aux"));
226
227         if (had_depfile) {
228                 LYXERR(Debug::DEPEND, "Dependency file exists");
229                 // Read the dep file:
230                 had_depfile = head.read(depfile);
231         }
232
233         if (had_depfile) {
234                 if (runparams.includeall) {
235                         // On an "includeall" call (whose purpose is to set up/maintain counters and references
236                         // for includeonly), we remove the master from the dependency list since
237                         // (1) it will be checked anyway on the subsequent includeonly run
238                         // (2) the master is always changed (due to the \includeonly line), and this alone would
239                         //     trigger a complete, expensive run each time
240                         head.remove_file(file);
241                         // Also remove all children which are included
242                         Buffer const * buf = theBufferList().getBufferFromTmp(file.absFileName());
243                         if (buf && buf->params().maintain_unincluded_children == BufferParams::CM_Mostly) {
244                                 for (auto const & incfile : buf->params().getIncludedChildren()) {
245                                         string const incm =
246                                                 DocFileName(changeExtension(makeAbsPath(incfile, path)
247                                                                             .absFileName(), ".tex")).mangledFileName();
248                                         FileName inc = makeAbsPath(incm, file.onlyPath().realPath());
249                                         head.remove_file(inc);
250                                 }
251                         }
252                 }
253                 // Update the checksums
254                 head.update();
255                 // Can't just check if anything has changed because it might
256                 // have aborted on error last time... in which cas we need
257                 // to re-run latex and collect the error messages
258                 // (even if they are the same).
259                 if (!output_file.exists()) {
260                         LYXERR(Debug::DEPEND,
261                                 "re-running LaTeX because output file doesn't exist.");
262                 } else if (!head.sumchange()) {
263                         LYXERR(Debug::DEPEND, "return no_change");
264                         return NO_CHANGE;
265                 } else {
266                         LYXERR(Debug::DEPEND, "Dependency file has changed");
267                 }
268
269                 if (head.extchanged(".bib") || head.extchanged(".bst"))
270                         run_bibtex = true;
271         } else
272                 LYXERR(Debug::DEPEND,
273                         "Dependency file does not exist, or has wrong format");
274
275         /// We scan the aux file even when had_depfile = false,
276         /// because we can run pdflatex on the file after running latex on it,
277         /// in which case we will not need to run bibtex again.
278         vector<AuxInfo> bibtex_info_old;
279         if (!run_bibtex)
280                 bibtex_info_old = scanAuxFiles(aux_file, runparams.only_childbibs);
281
282         ++count;
283         LYXERR(Debug::OUTFILE, "Run #" << count);
284         message(runMessage(count));
285
286         int exit_code = startscript();
287         if (exit_code == Systemcall::KILLED || exit_code == Systemcall::TIMEOUT)
288                 return exit_code;
289
290         scanres = scanLogFile(terr);
291         if (scanres & ERROR_RERUN) {
292                 LYXERR(Debug::OUTFILE, "Rerunning LaTeX");
293                 terr.clearErrors();
294                 exit_code = startscript();
295                 if (exit_code == Systemcall::KILLED || exit_code == Systemcall::TIMEOUT)
296                         return exit_code;
297                 scanres = scanLogFile(terr);
298         }
299
300         vector<AuxInfo> const bibtex_info = scanAuxFiles(aux_file, runparams.only_childbibs);
301         if (!run_bibtex && bibtex_info_old != bibtex_info)
302                 run_bibtex = true;
303
304         // update the dependencies.
305         deplog(head); // reads the latex log
306         head.update();
307
308         // 1
309         // At this point we must run the bibliography processor if needed.
310         // First, check if we're using biber instead of bibtex --
311         // biber writes no info to the aux file, so we just check
312         // if a bcf file exists (and if it was updated).
313         FileName const bcffile(changeExtension(file.absFileName(), ".bcf"));
314         biber |= head.exist(bcffile);
315
316         // If (scanres & UNDEF_CIT || scanres & RERUN || run_bibtex)
317         // We do not run bibtex/biber on an "includeall" call (whose purpose is
318         // to set up/maintain counters and references for includeonly) since
319         // (1) bibliographic references will be updated on the subsequent includeonly run
320         // (2) this would trigger a complete run each time (as references in non-included
321         //     children are removed on subsequent includeonly runs)
322         if (!runparams.includeall && (scanres & UNDEF_CIT || run_bibtex)) {
323                 // Here we must scan the .aux file and look for
324                 // "\bibdata" and/or "\bibstyle". If one of those
325                 // tags is found -> run bibtex and set rerun = true;
326                 // no checks for now
327                 LYXERR(Debug::OUTFILE, "Running Bibliography Processor.");
328                 message(_("Running Bibliography Processor."));
329                 updateBibtexDependencies(head, bibtex_info);
330                 int exit_code;
331                 rerun |= runBibTeX(bibtex_info, runparams, exit_code);
332                 if (exit_code == Systemcall::KILLED || exit_code == Systemcall::TIMEOUT)
333                         return exit_code;
334                 FileName const blgfile(changeExtension(file.absFileName(), ".blg"));
335                 if (blgfile.exists())
336                         bscanres = scanBlgFile(head, terr);
337         } else if (!had_depfile) {
338                 /// If we run pdflatex on the file after running latex on it,
339                 /// then we do not need to run bibtex, but we do need to
340                 /// insert the .bib and .bst files into the .dep-pdf file.
341                 updateBibtexDependencies(head, bibtex_info);
342         }
343
344         // 2
345         // We know on this point that latex has been run once (or we just
346         // returned) and the question now is to decide if we need to run
347         // it any more. This is done by asking if any of the files in the
348         // dependency file has changed. (remember that the checksum for
349         // a given file is reported to have changed if it just was created)
350         //     -> if changed or rerun == true:
351         //             run latex once more and
352         //             update the dependency structure
353         //     -> if not changed:
354         //             we do nothing at this point
355         //
356         if (rerun || head.sumchange()) {
357                 rerun = false;
358                 ++count;
359                 LYXERR(Debug::DEPEND, "Dep. file has changed or rerun requested");
360                 LYXERR(Debug::OUTFILE, "Run #" << count);
361                 message(runMessage(count));
362                 int exit_code = startscript();
363                 if (exit_code == Systemcall::KILLED || exit_code == Systemcall::TIMEOUT)
364                         return exit_code;
365                 scanres = scanLogFile(terr);
366
367                 // update the depedencies
368                 deplog(head); // reads the latex log
369                 head.update();
370         } else {
371                 LYXERR(Debug::DEPEND, "Dep. file has NOT changed");
372         }
373
374         // 3
375         // Rerun bibliography processor?
376         // Complex bibliography packages such as Biblatex require
377         // an additional bibtex cycle sometimes.
378         // We do not run bibtex/biber on an "includeall" call (whose purpose is
379         // to set up/maintain counters and references for includeonly) since
380         // (1) bibliographic references will be updated on the subsequent includeonly run
381         // (2) this would trigger a complete run each time (as references in non-included
382         //     children are removed on subsequent includeonly runs)
383         if (!runparams.includeall && scanres & UNDEF_CIT) {
384                 // Here we must scan the .aux file and look for
385                 // "\bibdata" and/or "\bibstyle". If one of those
386                 // tags is found -> run bibtex and set rerun = true;
387                 // no checks for now
388                 LYXERR(Debug::OUTFILE, "Re-Running Bibliography Processor.");
389                 message(_("Re-Running Bibliography Processor."));
390                 updateBibtexDependencies(head, bibtex_info);
391                 int exit_code;
392                 rerun |= runBibTeX(bibtex_info, runparams, exit_code);
393                 if (exit_code == Systemcall::KILLED || exit_code == Systemcall::TIMEOUT)
394                         return exit_code;
395                 FileName const blgfile(changeExtension(file.absFileName(), ".blg"));
396                 if (blgfile.exists())
397                         bscanres = scanBlgFile(head, terr);
398         }
399
400         // 4
401         // After the bibliography was processed, we need more passes of LaTeX
402         // in order to resolve the citations. We need to do this before the index
403         // is being generated (since we need the correct pagination, see #2696).
404         // Also, memoir (at least) writes an empty *idx file in the first place.
405         // A further latex run is needed in that case as well.
406         FileName const idxfile(changeExtension(file.absFileName(), ".idx"));
407         if (run_bibtex || (idxfile.exists() && idxfile.isFileEmpty())) {
408                 while ((head.sumchange() || rerun || (scanres & RERUN))
409                        && count < MAX_RUN) {
410                         // Yes rerun until message goes away, or until
411                         // MAX_RUNS are reached.
412                         rerun = false;
413                         ++count;
414                         LYXERR(Debug::OUTFILE, "Run #" << count);
415                         message(runMessage(count));
416                         startscript();
417                         scanres = scanLogFile(terr);
418         
419                         // keep this updated
420                         head.update();
421                 }
422         }
423
424         // 5
425         // Now that we have final pagination, run the index and nomencl processors
426         if (idxfile.exists()) {
427                 // no checks for now
428                 LYXERR(Debug::OUTFILE, "Running Index Processor.");
429                 message(_("Running Index Processor."));
430                 // onlyFileName() is needed for cygwin
431                 int const ret = 
432                                 runMakeIndex(onlyFileName(idxfile.absFileName()), runparams);
433                 if (ret == Systemcall::KILLED || ret == Systemcall::TIMEOUT)
434                         return ret;
435                 else if (ret != Systemcall::OK) {
436                         iscanres |= INDEX_ERROR;
437                         terr.insertError(0,
438                                          _("Index Processor Error"),
439                                          _("The index processor did not run successfully. "
440                                            "Please check the output of View > Messages Pane!"));
441                 }
442                 FileName const ilgfile(changeExtension(file.absFileName(), ".ilg"));
443                 if (ilgfile.exists())
444                         iscanres = scanIlgFile(terr);
445                 rerun = true;
446         }
447         FileName const nlofile(changeExtension(file.absFileName(), ".nlo"));
448         // If all nomencl entries are removed, nomencl writes an empty nlo file.
449         // DepTable::hasChanged() returns false in this case, since it does not
450         // distinguish empty files from non-existing files. This is why we need
451         // the extra checks here (to trigger a rerun). Cf. discussions in #8905.
452         // FIXME: Sort out the real problem in DepTable.
453         if (head.haschanged(nlofile) || (nlofile.exists() && nlofile.isFileEmpty())) {
454                 int const ret = runMakeIndexNomencl(file, ".nlo", ".nls");
455                 if (ret == Systemcall::KILLED || ret == Systemcall::TIMEOUT)
456                         return ret;
457                 rerun = true;
458         }
459         FileName const glofile(changeExtension(file.absFileName(), ".glo"));
460         if (head.haschanged(glofile)) {
461                 int const ret = runMakeIndexNomencl(file, ".glo", ".gls");
462                 if (ret)
463                         return ret;
464                 rerun = true;
465         }
466
467         // 6
468         // We will re-run latex if the log file asks for it,
469         // or if the sumchange() is true.
470         //     -> rerun asked for:
471         //             run latex and
472         //             remake the dependency file
473         //             goto 2 or return if max runs are reached.
474         //     -> rerun not asked for:
475         //             just return (fall out of bottom of func)
476         //
477         while ((head.sumchange() || rerun || (scanres & RERUN))
478                && count < MAX_RUN) {
479                 // Yes rerun until message goes away, or until
480                 // MAX_RUNS are reached.
481                 rerun = false;
482                 ++count;
483                 LYXERR(Debug::OUTFILE, "Run #" << count);
484                 message(runMessage(count));
485                 startscript();
486                 scanres = scanLogFile(terr);
487
488                 // keep this updated
489                 head.update();
490         }
491
492         // Write the dependencies to file.
493         head.write(depfile);
494
495         if (exit_code) {
496                 // add flag here, just before return, instead of when exit_code
497                 // is defined because scanres is sometimes overwritten above
498                 // (e.g. rerun)
499                 scanres |= NONZERO_ERROR;
500         }
501
502         LYXERR(Debug::OUTFILE, "Done.");
503
504         if (bscanres & ERRORS)
505                 return bscanres; // return on error
506
507         if (iscanres & ERRORS)
508                 return iscanres; // return on error
509
510         return scanres;
511 }
512
513
514 int LaTeX::startscript()
515 {
516         // onlyFileName() is needed for cygwin
517         string tmp = cmd + ' '
518                      + quoteName(onlyFileName(file.toFilesystemEncoding()))
519                      + " > " + os::nulldev();
520         Systemcall one;
521         Systemcall::Starttype const starttype = 
522                 allow_cancel ? Systemcall::WaitLoop : Systemcall::Wait;
523         return one.startscript(starttype, tmp, path, lpath, true);
524 }
525
526
527 int LaTeX::runMakeIndex(string const & f, OutputParams const & rp,
528                          string const & params)
529 {
530         string tmp = rp.use_japanese ?
531                 lyxrc.jindex_command : lyxrc.index_command;
532
533         if (!rp.index_command.empty())
534                 tmp = rp.index_command;
535
536         Language const * doc_lang = languages.getLanguage(rp.document_language);
537         
538         if (contains(tmp, "$$x")) {
539                 // This adds appropriate [te]xindy options
540                 // such as language and codepage (for the
541                 // main document language/encoding) as well
542                 // as input markup (latex or xelatex)
543                 string xdyopts = doc_lang ? doc_lang->xindy() : string();
544                 if (!xdyopts.empty())
545                         xdyopts = "-L " + xdyopts;
546                 if (rp.isFullUnicode() && rp.encoding->package() == Encoding::none) {
547                         if (!xdyopts.empty())
548                                 xdyopts += " ";
549                         // xelatex includes lualatex
550                         xdyopts += "-I xelatex";
551                 }
552                 else if (rp.encoding->iconvName() == "UTF-8") {
553                         if (!xdyopts.empty())
554                                 xdyopts += " ";
555                         // -I not really needed for texindy, but for xindy
556                         xdyopts += "-C utf8 -I latex";
557                 }
558                 else {
559                         if (!xdyopts.empty())
560                                 xdyopts += " ";
561                         // not really needed for texindy, but for xindy
562                         xdyopts += "-I latex";
563                 }
564                 tmp = subst(tmp, "$$x", xdyopts);
565         }
566
567         if (contains(tmp, "$$b")) {
568                 // advise xindy to write a log file
569                 tmp = subst(tmp, "$$b", removeExtension(f));
570         }
571
572         LYXERR(Debug::OUTFILE,
573                 "idx file has been made, running index processor ("
574                 << tmp << ") on file " << f);
575
576         if (doc_lang) {
577                 tmp = subst(tmp, "$$lang", doc_lang->babel());
578                 tmp = subst(tmp, "$$lcode", doc_lang->code());
579         }
580         if (rp.use_indices) {
581                 tmp = lyxrc.splitindex_command + " -m " + quoteName(tmp);
582                 LYXERR(Debug::OUTFILE,
583                 "Multiple indices. Using splitindex command: " << tmp);
584         }
585         tmp += ' ';
586         tmp += quoteName(f);
587         tmp += params;
588         Systemcall one;
589         Systemcall::Starttype const starttype = 
590                 allow_cancel ? Systemcall::WaitLoop : Systemcall::Wait;
591         return one.startscript(starttype, tmp, path, lpath, true);
592 }
593
594
595 int LaTeX::runMakeIndexNomencl(FileName const & fname,
596                 string const & nlo, string const & nls)
597 {
598         LYXERR(Debug::OUTFILE, "Running Nomenclature Processor.");
599         message(_("Running Nomenclature Processor."));
600         string tmp = lyxrc.nomencl_command + ' ';
601         // onlyFileName() is needed for cygwin
602         tmp += quoteName(onlyFileName(changeExtension(fname.absFileName(), nlo)));
603         tmp += " -o "
604                 + onlyFileName(changeExtension(fname.toFilesystemEncoding(), nls));
605         Systemcall one;
606         Systemcall::Starttype const starttype = 
607                 allow_cancel ? Systemcall::WaitLoop : Systemcall::Wait;
608         return one.startscript(starttype, tmp, path, lpath, true);
609 }
610
611
612 vector<AuxInfo> const
613 LaTeX::scanAuxFiles(FileName const & fname, bool const only_childbibs)
614 {
615         vector<AuxInfo> result;
616
617         // With chapterbib, we have to bibtex all children's aux files
618         // but _not_ the master's!
619         if (only_childbibs) {
620                 for (string const &s: children) {
621                         FileName fn =
622                                 makeAbsPath(s, fname.onlyPath().realPath());
623                         fn.changeExtension("aux");
624                         if (fn.exists())
625                                 result.push_back(scanAuxFile(fn));
626                 }
627                 return result;
628         }
629
630         result.push_back(scanAuxFile(fname));
631
632         // This is for bibtopic
633         string const basename = removeExtension(fname.absFileName());
634         for (int i = 1; i < 1000; ++i) {
635                 FileName const file2(basename
636                         + '.' + convert<string>(i)
637                         + ".aux");
638                 if (!file2.exists())
639                         break;
640                 result.push_back(scanAuxFile(file2));
641         }
642         return result;
643 }
644
645
646 AuxInfo const LaTeX::scanAuxFile(FileName const & fname)
647 {
648         AuxInfo result;
649         result.aux_file = fname;
650         scanAuxFile(fname, result);
651         return result;
652 }
653
654
655 void LaTeX::scanAuxFile(FileName const & fname, AuxInfo & aux_info)
656 {
657         LYXERR(Debug::OUTFILE, "Scanning aux file: " << fname);
658
659         ifstream ifs(fname.toFilesystemEncoding().c_str());
660         string token;
661         static regex const reg1("\\\\citation\\{([^}]+)\\}");
662         static regex const reg2("\\\\bibdata\\{([^}]+)\\}");
663         static regex const reg3("\\\\bibstyle\\{([^}]+)\\}");
664         static regex const reg4("\\\\@input\\{([^}]+)\\}");
665
666         while (getline(ifs, token)) {
667                 token = rtrim(token, "\r");
668                 smatch sub;
669                 // FIXME UNICODE: We assume that citation keys and filenames
670                 // in the aux file are in the file system encoding.
671                 token = to_utf8(from_filesystem8bit(token));
672                 if (regex_match(token, sub, reg1)) {
673                         string data = sub.str(1);
674                         while (!data.empty()) {
675                                 string citation;
676                                 data = split(data, citation, ',');
677                                 LYXERR(Debug::OUTFILE, "Citation: " << citation);
678                                 aux_info.citations.insert(citation);
679                         }
680                 } else if (regex_match(token, sub, reg2)) {
681                         string data = sub.str(1);
682                         // data is now all the bib files separated by ','
683                         // get them one by one and pass them to the helper
684                         while (!data.empty()) {
685                                 string database;
686                                 data = split(data, database, ',');
687                                 database = changeExtension(database, "bib");
688                                 LYXERR(Debug::OUTFILE, "BibTeX database: `" << database << '\'');
689                                 aux_info.databases.insert(database);
690                         }
691                 } else if (regex_match(token, sub, reg3)) {
692                         string style = sub.str(1);
693                         // token is now the style file
694                         // pass it to the helper
695                         style = changeExtension(style, "bst");
696                         LYXERR(Debug::OUTFILE, "BibTeX style: `" << style << '\'');
697                         aux_info.styles.insert(style);
698                 } else if (regex_match(token, sub, reg4)) {
699                         string const file2 = sub.str(1);
700                         scanAuxFile(makeAbsPath(file2), aux_info);
701                 }
702         }
703 }
704
705
706 void LaTeX::updateBibtexDependencies(DepTable & dep,
707                                      vector<AuxInfo> const & bibtex_info)
708 {
709         // Since a run of Bibtex mandates more latex runs it is ok to
710         // remove all ".bib" and ".bst" files.
711         dep.remove_files_with_extension(".bib");
712         dep.remove_files_with_extension(".bst");
713         //string aux = OnlyFileName(ChangeExtension(file, ".aux"));
714
715         for (vector<AuxInfo>::const_iterator it = bibtex_info.begin();
716              it != bibtex_info.end(); ++it) {
717                 for (set<string>::const_iterator it2 = it->databases.begin();
718                      it2 != it->databases.end(); ++it2) {
719                         FileName const file = findtexfile(*it2, "bib");
720                         if (!file.empty())
721                                 dep.insert(file, true);
722                 }
723
724                 for (set<string>::const_iterator it2 = it->styles.begin();
725                      it2 != it->styles.end(); ++it2) {
726                         FileName const file = findtexfile(*it2, "bst");
727                         if (!file.empty())
728                                 dep.insert(file, true);
729                 }
730         }
731
732         // biber writes nothing into the aux file.
733         // Instead, we have to scan the blg file
734         if (biber) {
735                 TeXErrors terr;
736                 scanBlgFile(dep, terr);
737         }
738 }
739
740
741 bool LaTeX::runBibTeX(vector<AuxInfo> const & bibtex_info,
742                       OutputParams const & rp, int & exit_code)
743 {
744         bool result = false;
745         exit_code = 0;
746         for (vector<AuxInfo>::const_iterator it = bibtex_info.begin();
747              it != bibtex_info.end(); ++it) {
748                 if (!biber && it->databases.empty())
749                         continue;
750                 result = true;
751
752                 string tmp = rp.bibtex_command;
753                 tmp += " ";
754                 // onlyFileName() is needed for cygwin
755                 tmp += quoteName(onlyFileName(removeExtension(
756                                 it->aux_file.absFileName())));
757                 Systemcall one;
758                 Systemcall::Starttype const starttype = 
759                         allow_cancel ? Systemcall::WaitLoop : Systemcall::Wait;
760                 exit_code = one.startscript(starttype, tmp, path, lpath, true);
761                 if (exit_code) {
762                         return result;
763                 }
764         }
765         // Return whether bibtex was run
766         return result;
767 }
768
769
770 //helper func for scanLogFile; gets line number X from strings "... on input line X ..."
771 //returns 0 if none is found
772 int getLineNumber(const string &token){
773         string l = support::token(token, ' ', tokenPos(token,' ',"line") + 1);
774         return l.empty() ? 0 : convert<int>(l);
775 }
776
777
778 int LaTeX::scanLogFile(TeXErrors & terr)
779 {
780         int last_line = -1;
781         int line_count = 1;
782         int retval = NO_ERRORS;
783         string tmp =
784                 onlyFileName(changeExtension(file.absFileName(), ".log"));
785         LYXERR(Debug::OUTFILE, "Log file: " << tmp);
786         FileName const fn = makeAbsPath(tmp);
787         // FIXME we should use an ifdocstream here and a docstring for token
788         // below. The encoding of the log file depends on the _output_ (font)
789         // encoding of the TeX file (T1, TU etc.). See #10728.
790         ifstream ifs(fn.toFilesystemEncoding().c_str());
791         bool fle_style = false;
792         static regex const file_line_error(".+\\.\\D+:[0-9]+: (.+)");
793         static regex const child_file("[^0-9]*([0-9]+[A-Za-z]*_.+\\.tex).*");
794         static regex const undef_ref(".*Reference `(\\w+)\\' on page.*");
795         // Flag for 'File ended while scanning' message.
796         // We need to wait for subsequent processing.
797         string wait_for_error;
798         string child_name;
799         int pnest = 0;
800         stack <pair<string, int> > child;
801         children.clear();
802
803         terr.clearRefs();
804
805         string token;
806         string ml_token;
807         while (getline(ifs, token)) {
808                 // MikTeX sometimes inserts \0 in the log file. They can't be
809                 // removed directly with the existing string utility
810                 // functions, so convert them first to \r, and remove all
811                 // \r's afterwards, since we need to remove them anyway.
812                 token = subst(token, '\0', '\r');
813                 token = subst(token, "\r", "");
814                 smatch sub;
815
816                 LYXERR(Debug::OUTFILE, "Log line: " << token);
817
818                 if (token.empty())
819                         continue;
820
821                 if (!ml_token.empty())
822                         ml_token += token;
823
824                 // Track child documents
825                 for (size_t i = 0; i < token.length(); ++i) {
826                         if (token[i] == '(') {
827                                 ++pnest;
828                                 size_t j = token.find('(', i + 1);
829                                 size_t len = j == string::npos
830                                                 ? token.substr(i + 1).length()
831                                                 : j - i - 1;
832                                 string const substr = token.substr(i + 1, len);
833                                 if (regex_match(substr, sub, child_file)) {
834                                         string const name = sub.str(1);
835                                         // Sometimes also masters have a name that matches
836                                         // (if their name starts with a number and _)
837                                         if (name != file.onlyFileName()) {
838                                                 child.push(make_pair(name, pnest));
839                                                 children.push_back(name);
840                                         }
841                                         i += len;
842                                 }
843                         } else if (token[i] == ')') {
844                                 if (!child.empty()
845                                     && child.top().second == pnest)
846                                         child.pop();
847                                 --pnest;
848                         }
849                 }
850                 child_name = child.empty() ? empty_string() : child.top().first;
851
852                 if (contains(token, "file:line:error style messages enabled"))
853                         fle_style = true;
854
855                 //Handles both "LaTeX Warning:" & "Package natbib Warning:"
856                 //Various handlers for missing citations below won't catch the problem if citation
857                 //key is long (>~25chars), because pdflatex splits output at line length 80.
858                 //TODO: TL 2020 engines will contain new commandline switch --cnf-line which we  
859                 //can use to set max_print_line variable for appropriate length and detect all
860                 //errors correctly.
861                 if (!runparams.includeall && (contains(token, "There were undefined citations.") ||
862                     prefixIs(token, "Package biblatex Warning: The following entry could not be found")))
863                         retval |= UNDEF_CIT;
864
865                 if (prefixIs(token, "LaTeX Warning:")
866                     || prefixIs(token, "! pdfTeX warning")
867                     || prefixIs(ml_token, "LaTeX Warning:")
868                     || prefixIs(ml_token, "! pdfTeX warning")) {
869                         // Here shall we handle different
870                         // types of warnings
871                         retval |= LATEX_WARNING;
872                         LYXERR(Debug::OUTFILE, "LaTeX Warning.");
873                         if (contains(token, "Rerun to get cross-references")) {
874                                 retval |= RERUN;
875                                 LYXERR(Debug::OUTFILE, "We should rerun.");
876                         // package clefval needs 2 latex runs before bibtex
877                         } else if (contains(token, "Value of")
878                                    && contains(token, "on page")
879                                    && contains(token, "undefined")) {
880                                 retval |= ERROR_RERUN;
881                                 LYXERR(Debug::OUTFILE, "Force rerun.");
882                         // package etaremune
883                         } else if (contains(token, "Etaremune labels have changed")) {
884                                 retval |= ERROR_RERUN;
885                                 LYXERR(Debug::OUTFILE, "Force rerun.");
886                         // package enotez
887                         } else if (contains(token, "Endnotes may have changed. Rerun")) {
888                                 retval |= RERUN;
889                                 LYXERR(Debug::OUTFILE, "We should rerun.");
890                         //"Citation `cit' on page X undefined on input line X."
891                         } else if (!runparams.includeall && contains(token, "Citation")
892                                    //&& contains(token, "on input line") //often split to newline
893                                    && contains(token, "undefined")) {
894                                 retval |= UNDEF_CIT;
895                                 terr.insertRef(getLineNumber(token), from_ascii("Citation undefined"),
896                                         from_utf8(token), child_name);
897                         //"Reference `X' on page Y undefined on input line Z."
898                         // This warning might be broken accross multiple lines with long labels.
899                         // Thus we check that
900                         } else if (contains(token, "Reference `") && !contains(token, "on input line")) {
901                                 // Rest of warning in next line(s)
902                                 // Save to ml_token
903                                 ml_token = token;
904                         } else if (!ml_token.empty() && contains(ml_token, "Reference `")
905                                    && !contains(ml_token, "on input line")) {
906                                 // not finished yet. Continue with next line.
907                                 continue;
908                         } else if (!ml_token.empty() && contains(ml_token, "Reference `")
909                                    && contains(ml_token, "on input line")) {
910                                 // We have collected the whole warning now.
911                                 if (!contains(ml_token, "undefined")) {
912                                         // Not the warning we are looking for
913                                         ml_token.clear();
914                                         continue;
915                                 }
916                                 if (regex_match(ml_token, sub, undef_ref)) {
917                                         string const ref = sub.str(1);
918                                         Buffer const * buf = theBufferList().getBufferFromTmp(file.absFileName());
919                                         if (!buf || !buf->masterBuffer()->activeLabel(from_utf8(ref))) {
920                                                 terr.insertRef(getLineNumber(ml_token), from_ascii("Reference undefined"),
921                                                         from_utf8(ml_token), child_name);
922                                                 retval |= UNDEF_UNKNOWN_REF;
923                                         }
924                                 }
925                                 ml_token.clear();
926                                 retval |= UNDEF_REF;
927                         } else if (contains(token, "Reference `")
928                                    && contains(token, "on input line")
929                                    && contains(token, "undefined")) {
930                                 if (regex_match(token, sub, undef_ref)) {
931                                         string const ref = sub.str(1);
932                                         Buffer const * buf = theBufferList().getBufferFromTmp(file.absFileName());
933                                         if (!buf || !buf->masterBuffer()->activeLabel(from_utf8(ref))) {
934                                                 terr.insertRef(getLineNumber(token), from_ascii("Reference undefined"),
935                                                         from_utf8(token), child_name);
936                                                 retval |= UNDEF_UNKNOWN_REF;
937                                         }
938                                 }
939                                 retval |= UNDEF_REF;
940                         // In case the above checks fail we catch at least this generic statement
941                         // occuring for both CIT & REF.
942                         } else if (!runparams.includeall && contains(token, "There were undefined references.")) {
943                                 if (!(retval & UNDEF_CIT)) //if not handled already
944                                         retval |= UNDEF_REF;
945                         }
946
947                 } else if (prefixIs(token, "Package")) {
948                         // Package warnings
949                         retval |= PACKAGE_WARNING;
950                         if (contains(token, "natbib Warning:")) {
951                                 // Natbib warnings
952                                 if (!runparams.includeall
953                                     && contains(token, "Citation")
954                                     && contains(token, "on page")
955                                     && contains(token, "undefined")) {
956                                         retval |= UNDEF_CIT;
957                                         //Unf only keys up to ~6 chars will make it due to line splits
958                                         terr.insertRef(getLineNumber(token), from_ascii("Citation undefined"),
959                                                 from_utf8(token), child_name);
960                                 }
961                         } else if (!runparams.includeall && contains(token, "run BibTeX")) {
962                                 retval |= UNDEF_CIT;
963                         } else if (!runparams.includeall && contains(token, "run Biber")) {
964                                 retval |= UNDEF_CIT;
965                                 biber = true;
966                         } else if (contains(token, "Rerun LaTeX") ||
967                                    contains(token, "Please rerun LaTeX") ||
968                                    contains(token, "Rerun to get")) {
969                                 // at least longtable.sty and bibtopic.sty
970                                 // might use this.
971                                 LYXERR(Debug::OUTFILE, "We should rerun.");
972                                 retval |= RERUN;
973                         }
974                 } else if (prefixIs(token, "LETTRE WARNING:")) {
975                         if (contains(token, "veuillez recompiler")) {
976                                 // lettre.cls
977                                 LYXERR(Debug::OUTFILE, "We should rerun.");
978                                 retval |= RERUN;
979                         }
980                 } else if (token[0] == '(') {
981                         if (contains(token, "Rerun LaTeX") ||
982                             contains(token, "Rerun to get")) {
983                                 // Used by natbib
984                                 LYXERR(Debug::OUTFILE, "We should rerun.");
985                                 retval |= RERUN;
986                         }
987                 } else if (prefixIs(token, "! ")
988                             || (fle_style
989                                 && regex_match(token, sub, file_line_error)
990                                 && !contains(token, "pdfTeX warning"))) {
991                            // Ok, we have something that looks like a TeX Error
992                            // but what do we really have.
993
994                         // Just get the error description:
995                         string desc;
996                         if (prefixIs(token, "! "))
997                                 desc = string(token, 2);
998                         else if (fle_style)
999                                 desc = sub.str();
1000                         if (contains(token, "LaTeX Error:"))
1001                                 retval |= LATEX_ERROR;
1002
1003                         if (prefixIs(token, "! File ended while scanning")) {
1004                                 if (prefixIs(token, "! File ended while scanning use of \\Hy@setref@link.")){
1005                                         // bug 7344. We must rerun LaTeX if hyperref has been toggled.
1006                                         retval |= ERROR_RERUN;
1007                                         LYXERR(Debug::OUTFILE, "Force rerun.");
1008                                 } else {
1009                                         // bug 6445. At this point its not clear we finish with error.
1010                                         wait_for_error = desc;
1011                                         continue;
1012                                 }
1013                         }
1014
1015                         if (prefixIs(token, "! Incomplete \\if")) {
1016                                 // bug 10666. At this point its not clear we finish with error.
1017                                 wait_for_error = desc;
1018                                 continue;
1019                         }
1020
1021                         if (prefixIs(token, "! Paragraph ended before \\Hy@setref@link was complete.")){
1022                                         // bug 7344. We must rerun LaTeX if hyperref has been toggled.
1023                                         retval |= ERROR_RERUN;
1024                                         LYXERR(Debug::OUTFILE, "Force rerun.");
1025                         }
1026
1027                         if (!wait_for_error.empty() && prefixIs(token, "! Emergency stop.")){
1028                                 retval |= LATEX_ERROR;
1029                                 string errstr;
1030                                 int count = 0;
1031                                 errstr = wait_for_error;
1032                                 wait_for_error.clear();
1033                                 do {
1034                                         if (!getline(ifs, tmp))
1035                                                 break;
1036                                         tmp = rtrim(tmp, "\r");
1037                                         errstr += "\n" + tmp;
1038                                         if (++count > 5)
1039                                                 break;
1040                                 } while (!contains(tmp, "(job aborted"));
1041
1042                                 terr.insertError(0,
1043                                                  from_ascii("Emergency stop"),
1044                                                  from_local8bit(errstr),
1045                                                  child_name);
1046                         }
1047
1048                         // get the next line
1049                         int count = 0;
1050                         // We also collect intermediate lines
1051                         // This is needed for errors in preamble
1052                         string intermediate;
1053                         do {
1054                                 if (!getline(ifs, tmp))
1055                                         break;
1056                                 tmp = rtrim(tmp, "\r");
1057                                 if (!prefixIs(tmp, "l."))
1058                                         intermediate += tmp;
1059                                 // 15 is somewhat arbitrarily chosen, based on practice.
1060                                 // We used 10 for 14 years and increased it to 15 when we
1061                                 // saw one case.
1062                                 if (++count > 15)
1063                                         break;
1064                         } while (!prefixIs(tmp, "l."));
1065                         if (prefixIs(tmp, "l.")) {
1066                                 // we have a latex error
1067                                 retval |=  TEX_ERROR;
1068                                 if (contains(desc,
1069                                         "Package babel Error: You haven't defined the language")
1070                                     || contains(desc,
1071                                         "Package babel Error: You haven't loaded the option")
1072                                     || contains(desc,
1073                                         "Package babel Error: Unknown language"))
1074                                         retval |= ERROR_RERUN;
1075                                 // get the line number:
1076                                 int line = 0;
1077                                 sscanf(tmp.c_str(), "l.%d", &line);
1078                                 // get the rest of the message:
1079                                 string errstr(tmp, tmp.find(' '));
1080                                 bool preamble_error = false;
1081                                 if (suffixIs(errstr, "\\begin{document}")) {
1082                                         // this is an error in preamble
1083                                         // the real error is in the
1084                                         // intermediate lines
1085                                         errstr = intermediate;
1086                                         tmp = intermediate;
1087                                         preamble_error = true;
1088                                 }
1089                                 errstr += '\n';
1090                                 getline(ifs, tmp);
1091                                 tmp = rtrim(tmp, "\r");
1092                                 while (!contains(errstr, "l.")
1093                                        && !tmp.empty()
1094                                        && !prefixIs(tmp, "! ")
1095                                        && !contains(tmp, "(job aborted")) {
1096                                         errstr += tmp;
1097                                         errstr += "\n";
1098                                         getline(ifs, tmp);
1099                                         tmp = rtrim(tmp, "\r");
1100                                 }
1101                                 if (preamble_error)
1102                                         // Add a note that the error is to be found in preamble
1103                                         errstr += "\n" + to_utf8(_("(NOTE: The erroneous command is in the preamble)"));
1104                                 LYXERR(Debug::OUTFILE, "line: " << line << '\n'
1105                                                                 << "Desc: " << desc << '\n' << "Text: " << errstr);
1106                                 if (line == last_line)
1107                                         ++line_count;
1108                                 else {
1109                                         line_count = 1;
1110                                         last_line = line;
1111                                 }
1112                                 if (line_count <= 5) {
1113                                         // FIXME UNICODE
1114                                         // We have no idea what the encoding of
1115                                         // the log file is.
1116                                         // It seems that the output from the
1117                                         // latex compiler itself is pure ASCII,
1118                                         // but it can include bits from the
1119                                         // document, so whatever encoding we
1120                                         // assume here it can be wrong.
1121                                         terr.insertError(line,
1122                                                          from_local8bit(desc),
1123                                                          from_local8bit(errstr),
1124                                                          child_name);
1125                                         ++num_errors;
1126                                 }
1127                         }
1128                 } else {
1129                         // information messages, TeX warnings and other
1130                         // warnings we have not caught earlier.
1131                         if (prefixIs(token, "Overfull ")) {
1132                                 retval |= TEX_WARNING;
1133                         } else if (prefixIs(token, "Underfull ")) {
1134                                 retval |= TEX_WARNING;
1135                         } else if (!runparams.includeall && contains(token, "Rerun to get citations")) {
1136                                 // Natbib seems to use this.
1137                                 retval |= UNDEF_CIT;
1138                         } else if (contains(token, "No pages of output")
1139                                 || contains(token, "no pages of output")) {
1140                                 // No output file (e.g. the DVI or PDF) was created
1141                                 retval |= NO_OUTPUT;
1142                         } else if (contains(token, "Error 256 (driver return code)")) {
1143                                 // This is a xdvipdfmx driver error reported by XeTeX.
1144                                 // We have to check whether an output PDF file was created.
1145                                 FileName pdffile = file;
1146                                 pdffile.changeExtension("pdf");
1147                                 if (!pdffile.exists())
1148                                         // No output PDF file was created (see #10076)
1149                                         retval |= NO_OUTPUT;
1150                         } else if (contains(token, "That makes 100 errors")) {
1151                                 // More than 100 errors were reported
1152                                 retval |= TOO_MANY_ERRORS;
1153                         } else if (prefixIs(token, "!pdfTeX error:")) {
1154                                 // otherwise we dont catch e.g.:
1155                                 // !pdfTeX error: pdflatex (file feyn10): Font feyn10 at 600 not found
1156                                 retval |= ERRORS;
1157                                 terr.insertError(0,
1158                                                  from_ascii("pdfTeX Error"),
1159                                                  from_local8bit(token),
1160                                                  child_name);
1161                         } else if (!ignore_missing_glyphs
1162                                    && prefixIs(token, "Missing character: There is no ")
1163                                    && !contains(token, "nullfont")) {
1164                                 // Warning about missing glyph in selected font
1165                                 // may be dataloss (bug 9610)
1166                                 // but can be ignored for 'nullfont' (bug 10394).
1167                                 // as well as for ZERO WIDTH NON-JOINER (0x200C) which is
1168                                 // missing in many fonts and output for ligature break (bug 10727).
1169                                 // Since this error only occurs with utf8 output, we can safely assume
1170                                 // that the log file is utf8-encoded
1171                                 docstring const utoken = from_utf8(token);
1172                                 if (!contains(utoken, 0x200C)) {
1173                                         retval |= LATEX_ERROR;
1174                                         terr.insertError(0,
1175                                                          from_ascii("Missing glyphs!"),
1176                                                          utoken,
1177                                                          child_name);
1178                                 }
1179                         } else if (!wait_for_error.empty()) {
1180                                 // We collect information until we know we have an error.
1181                                 wait_for_error += token + '\n';
1182                         }
1183                 }
1184         }
1185         LYXERR(Debug::OUTFILE, "Log line: " << token);
1186         return retval;
1187 }
1188
1189
1190 namespace {
1191
1192 bool insertIfExists(FileName const & absname, DepTable & head)
1193 {
1194         if (absname.exists() && !absname.isDirectory()) {
1195                 head.insert(absname, true);
1196                 return true;
1197         }
1198         return false;
1199 }
1200
1201
1202 bool handleFoundFile(string const & ff, DepTable & head)
1203 {
1204         // convert from native os path to unix path
1205         string foundfile = os::internal_path(trim(ff));
1206
1207         LYXERR(Debug::DEPEND, "Found file: " << foundfile);
1208
1209         // Ok now we found a file.
1210         // Now we should make sure that this is a file that we can
1211         // access through the normal paths.
1212         // We will not try any fancy search methods to
1213         // find the file.
1214
1215         // (1) foundfile is an
1216         //     absolute path and should
1217         //     be inserted.
1218         FileName absname;
1219         if (FileName::isAbsolute(foundfile)) {
1220                 LYXERR(Debug::DEPEND, "AbsolutePath file: " << foundfile);
1221                 // On initial insert we want to do the update at once
1222                 // since this file cannot be a file generated by
1223                 // the latex run.
1224                 absname.set(foundfile);
1225                 if (!insertIfExists(absname, head)) {
1226                         // check for spaces
1227                         string strippedfile = foundfile;
1228                         while (contains(strippedfile, " ")) {
1229                                 // files with spaces are often enclosed in quotation
1230                                 // marks; those have to be removed
1231                                 string unquoted = subst(strippedfile, "\"", "");
1232                                 absname.set(unquoted);
1233                                 if (insertIfExists(absname, head))
1234                                         return true;
1235                                 // strip off part after last space and try again
1236                                 string tmp = strippedfile;
1237                                 rsplit(tmp, strippedfile, ' ');
1238                                 absname.set(strippedfile);
1239                                 if (insertIfExists(absname, head))
1240                                         return true;
1241                         }
1242                 }
1243         }
1244
1245         string onlyfile = onlyFileName(foundfile);
1246         absname = makeAbsPath(onlyfile);
1247
1248         // check for spaces
1249         while (contains(foundfile, ' ')) {
1250                 if (absname.exists())
1251                         // everything o.k.
1252                         break;
1253                 else {
1254                         // files with spaces are often enclosed in quotation
1255                         // marks; those have to be removed
1256                         string unquoted = subst(foundfile, "\"", "");
1257                         absname = makeAbsPath(unquoted);
1258                         if (absname.exists())
1259                                 break;
1260                         // strip off part after last space and try again
1261                         string strippedfile;
1262                         rsplit(foundfile, strippedfile, ' ');
1263                         foundfile = strippedfile;
1264                         onlyfile = onlyFileName(strippedfile);
1265                         absname = makeAbsPath(onlyfile);
1266                 }
1267         }
1268
1269         // (2) foundfile is in the tmpdir
1270         //     insert it into head
1271         if (absname.exists() && !absname.isDirectory()) {
1272                 // FIXME: This regex contained glo, but glo is used by the old
1273                 // version of nomencl.sty. Do we need to put it back?
1274                 static regex const unwanted("^.*\\.(aux|log|dvi|bbl|ind)$");
1275                 if (regex_match(onlyfile, unwanted)) {
1276                         LYXERR(Debug::DEPEND, "We don't want " << onlyfile
1277                                 << " in the dep file");
1278                 } else if (suffixIs(onlyfile, ".tex")) {
1279                         // This is a tex file generated by LyX
1280                         // and latex is not likely to change this
1281                         // during its runs.
1282                         LYXERR(Debug::DEPEND, "Tmpdir TeX file: " << onlyfile);
1283                         head.insert(absname, true);
1284                 } else {
1285                         LYXERR(Debug::DEPEND, "In tmpdir file:" << onlyfile);
1286                         head.insert(absname);
1287                 }
1288                 return true;
1289         } else {
1290                 LYXERR(Debug::DEPEND, "Not a file or we are unable to find it.");
1291                 return false;
1292         }
1293 }
1294
1295
1296 bool completeFilename(string const & ff, DepTable & head)
1297 {
1298         // If we do not find a dot, we suspect
1299         // a fragmental file name
1300         if (!contains(ff, '.'))
1301                 return false;
1302
1303         // if we have a dot, we let handleFoundFile decide
1304         return handleFoundFile(ff, head);
1305 }
1306
1307
1308 int iterateLine(string const & token, regex const & reg, string const & opening,
1309                 string const & closing, int fragment_pos, DepTable & head)
1310 {
1311         smatch what;
1312         string::const_iterator first = token.begin();
1313         string::const_iterator end = token.end();
1314         bool fragment = false;
1315         string last_match;
1316
1317         while (regex_search(first, end, what, reg)) {
1318                 // if we have a dot, try to handle as file
1319                 if (contains(what.str(1), '.')) {
1320                         first = what[0].second;
1321                         if (what.str(2) == closing) {
1322                                 handleFoundFile(what.str(1), head);
1323                                 // since we had a closing bracket,
1324                                 // do not investigate further
1325                                 fragment = false;
1326                         } else if (what.str(2) == opening) {
1327                                 // if we have another opening bracket,
1328                                 // we might have a nested file chain
1329                                 // as is (file.ext (subfile.ext))
1330                                 fragment = !handleFoundFile(rtrim(what.str(1)), head);
1331                                 // decrease first position by one in order to
1332                                 // consider the opening delimiter on next iteration
1333                                 if (first > token.begin())
1334                                         --first;
1335                         } else
1336                                 // if we have no closing bracket,
1337                                 // try to handle as file nevertheless
1338                                 fragment = !handleFoundFile(
1339                                         what.str(1) + what.str(2), head);
1340                 }
1341                 // if we do not have a dot, check if the line has
1342                 // a closing bracket (else, we suspect a line break)
1343                 else if (what.str(2) != closing) {
1344                         first = what[0].second;
1345                         fragment = true;
1346                 } else {
1347                         // we have a closing bracket, so the content
1348                         // is not a file name.
1349                         // no need to investigate further
1350                         first = what[0].second;
1351                         fragment = false;
1352                 }
1353                 last_match = what.str(1);
1354         }
1355
1356         // We need to consider the result from previous line iterations:
1357         // We might not find a fragment here, but another one might follow
1358         // E.g.: (filename.ext) <filenam
1359         // Vice versa, we consider the search completed if a real match
1360         // follows a potential fragment from a previous iteration.
1361         // E.g. <some text we considered a fragment (filename.ext)
1362         // result = -1 means we did not find a fragment!
1363         int result = -1;
1364         int last_match_pos = -1;
1365         if (!last_match.empty() && token.find(last_match) != string::npos)
1366                 last_match_pos = int(token.find(last_match));
1367         if (fragment) {
1368                 if (last_match_pos > fragment_pos)
1369                         result = last_match_pos;
1370                 else
1371                         result = fragment_pos;
1372         } else
1373                 if (last_match_pos < fragment_pos)
1374                         result = fragment_pos;
1375
1376         return result;
1377 }
1378
1379 } // namespace
1380
1381
1382 void LaTeX::deplog(DepTable & head)
1383 {
1384         // This function reads the LaTeX log file end extracts all the
1385         // external files used by the LaTeX run. The files are then
1386         // entered into the dependency file.
1387
1388         string const logfile =
1389                 onlyFileName(changeExtension(file.absFileName(), ".log"));
1390
1391         static regex const reg1("File: (.+).*");
1392         static regex const reg2("No file (.+)(.).*");
1393         static regex const reg3a("\\\\openout[0-9]+.*=.*`(.+)(..).*");
1394         // LuaTeX has a slightly different output
1395         static regex const reg3b("\\\\openout[0-9]+.*=\\s*(.+)");
1396         // If an index should be created, MikTex does not write a line like
1397         //    \openout# = 'sample.idx'.
1398         // but instead only a line like this into the log:
1399         //   Writing index file sample.idx
1400         static regex const reg4("Writing index file (.+).*");
1401         static regex const regoldnomencl("Writing glossary file (.+).*");
1402         static regex const regnomencl(".*Writing nomenclature file (.+).*");
1403         // If a toc should be created, MikTex does not write a line like
1404         //    \openout# = `sample.toc'.
1405         // but only a line like this into the log:
1406         //    \tf@toc=\write#
1407         // This line is also written by tetex.
1408         // This line is not present if no toc should be created.
1409         static regex const miktexTocReg("\\\\tf@toc=\\\\write.*");
1410         // file names can be enclosed in <...> (anywhere on the line)
1411         static regex const reg5(".*<[^>]+.*");
1412         // and also (...) anywhere on the line
1413         static regex const reg6(".*\\([^)]+.*");
1414
1415         FileName const fn = makeAbsPath(logfile);
1416         ifstream ifs(fn.toFilesystemEncoding().c_str());
1417         string lastline;
1418         while (ifs) {
1419                 // Ok, the scanning of files here is not sufficient.
1420                 // Sometimes files are named by "File: xxx" only
1421                 // Therefore we use some regexps to find files instead.
1422                 // Note: all file names and paths might contains spaces.
1423                 // Also, file names might be broken across lines. Therefore
1424                 // we mark (potential) fragments and merge those lines.
1425                 bool fragment = false;
1426                 string token;
1427                 getline(ifs, token);
1428                 // MikTeX sometimes inserts \0 in the log file. They can't be
1429                 // removed directly with the existing string utility
1430                 // functions, so convert them first to \r, and remove all
1431                 // \r's afterwards, since we need to remove them anyway.
1432                 token = subst(token, '\0', '\r');
1433                 token = subst(token, "\r", "");
1434                 if (token.empty() || token == ")") {
1435                         lastline = string();
1436                         continue;
1437                 }
1438
1439                 // FIXME UNICODE: We assume that the file names in the log
1440                 // file are in the file system encoding.
1441                 token = to_utf8(from_filesystem8bit(token));
1442
1443                 // Sometimes, filenames are broken across lines.
1444                 // We care for that and save suspicious lines.
1445                 // Here we exclude some cases where we are sure
1446                 // that there is no continued filename
1447                 if (!lastline.empty()) {
1448                         static regex const package_info("Package \\w+ Info: .*");
1449                         static regex const package_warning("Package \\w+ Warning: .*");
1450                         if (prefixIs(token, "File:") || prefixIs(token, "(Font)")
1451                             || prefixIs(token, "Package:")
1452                             || prefixIs(token, "Language:")
1453                             || prefixIs(token, "LaTeX Info:")
1454                             || prefixIs(token, "LaTeX Font Info:")
1455                             || prefixIs(token, "\\openout[")
1456                             || prefixIs(token, "))")
1457                             || regex_match(token, package_info)
1458                             || regex_match(token, package_warning))
1459                                 lastline = string();
1460                 }
1461
1462                 if (!lastline.empty())
1463                         // probably a continued filename from last line
1464                         token = lastline + token;
1465                 if (token.length() > 255) {
1466                         // string too long. Cut off.
1467                         token.erase(0, token.length() - 251);
1468                 }
1469
1470                 smatch sub;
1471
1472                 // (1) "File: file.ext"
1473                 if (regex_match(token, sub, reg1)) {
1474                         // is this a fragmental file name?
1475                         fragment = !completeFilename(sub.str(1), head);
1476                         // However, ...
1477                         if (suffixIs(token, ")"))
1478                                 // no fragment for sure
1479                                 fragment = false;
1480                 // (2) "No file file.ext"
1481                 } else if (regex_match(token, sub, reg2)) {
1482                         // file names must contains a dot, line ends with dot
1483                         if (contains(sub.str(1), '.') && sub.str(2) == ".")
1484                                 fragment = !handleFoundFile(sub.str(1), head);
1485                         else
1486                                 // we suspect a line break
1487                                 fragment = true;
1488                 // (3)(a) "\openout<nr> = `file.ext'."
1489                 } else if (regex_match(token, sub, reg3a)) {
1490                         // search for closing '. at the end of the line
1491                         if (sub.str(2) == "\'.")
1492                                 fragment = !handleFoundFile(sub.str(1), head);
1493                         else
1494                                 // potential fragment
1495                                 fragment = true;
1496                 // (3)(b) "\openout<nr> = file.ext" (LuaTeX)
1497                 } else if (regex_match(token, sub, reg3b)) {
1498                         // file names must contains a dot
1499                         if (contains(sub.str(1), '.'))
1500                                 fragment = !handleFoundFile(sub.str(1), head);
1501                         else
1502                                 // potential fragment
1503                                 fragment = true;
1504                 // (4) "Writing index file file.ext"
1505                 } else if (regex_match(token, sub, reg4))
1506                         // fragmential file name?
1507                         fragment = !completeFilename(sub.str(1), head);
1508                 // (5) "Writing nomenclature file file.ext"
1509                 else if (regex_match(token, sub, regnomencl) ||
1510                            regex_match(token, sub, regoldnomencl))
1511                         // fragmental file name?
1512                         fragment= !completeFilename(sub.str(1), head);
1513                 // (6) "\tf@toc=\write<nr>" (for MikTeX)
1514                 else if (regex_match(token, sub, miktexTocReg))
1515                         fragment = !handleFoundFile(onlyFileName(changeExtension(
1516                                                 file.absFileName(), ".toc")), head);
1517                 else
1518                         // not found, but we won't check further
1519                         fragment = false;
1520
1521                 int fragment_pos = -1;
1522                 // (7) "<file.ext>"
1523                 // We can have several of these on one line
1524                 // (and in addition to those above)
1525                 if (regex_match(token, sub, reg5)) {
1526                         // search for strings in <...>
1527                         static regex const reg5_1("<([^>]+)(.)");
1528                         fragment_pos = iterateLine(token, reg5_1, "<", ">",
1529                                                    fragment_pos, head);
1530                         fragment = (fragment_pos != -1);
1531                 }
1532
1533                 // (8) "(file.ext)"
1534                 // We can have several of these on one line
1535                 // this must be queried separated, because of
1536                 // cases such as "File: file.ext (type eps)"
1537                 // where "File: file.ext" would be skipped
1538                 if (regex_match(token, sub, reg6)) {
1539                         // search for strings in (...)
1540                         static regex const reg6_1("\\(([^()]+)(.)");
1541                         fragment_pos = iterateLine(token, reg6_1, "(", ")",
1542                                                    fragment_pos, head);
1543                         fragment = (fragment_pos != -1);
1544                 }
1545
1546                 if (fragment)
1547                         // probable linebreak within file name:
1548                         // save this line
1549                         lastline = token;
1550                 else
1551                         // no linebreak: reset
1552                         lastline = string();
1553         }
1554
1555         // Make sure that the main .tex file is in the dependency file.
1556         head.insert(file, true);
1557 }
1558
1559
1560 int LaTeX::scanBlgFile(DepTable & dep, TeXErrors & terr)
1561 {
1562         FileName const blg_file(changeExtension(file.absFileName(), "blg"));
1563         LYXERR(Debug::OUTFILE, "Scanning blg file: " << blg_file);
1564
1565         ifstream ifs(blg_file.toFilesystemEncoding().c_str());
1566         string token;
1567         static regex const reg1(".*Found (bibtex|BibTeX) data (file|source) '([^']+).*");
1568         static regex const bibtexError("^(.*---line [0-9]+ of file).*$");
1569         static regex const bibtexError2("^(.*---while reading file).*$");
1570         static regex const bibtexError3("(A bad cross reference---).*");
1571         static regex const bibtexError4("(Sorry---you've exceeded BibTeX's).*");
1572         static regex const bibtexError5("\\*Please notify the BibTeX maintainer\\*");
1573         static regex const biberError("^.*> (FATAL|ERROR) - (.*)$");
1574         int retval = NO_ERRORS;
1575
1576         string prevtoken;
1577         while (getline(ifs, token)) {
1578                 token = rtrim(token, "\r");
1579                 smatch sub;
1580                 // FIXME UNICODE: We assume that citation keys and filenames
1581                 // in the aux file are in the file system encoding.
1582                 token = to_utf8(from_filesystem8bit(token));
1583                 if (regex_match(token, sub, reg1)) {
1584                         string data = sub.str(3);
1585                         if (!data.empty()) {
1586                                 LYXERR(Debug::OUTFILE, "Found bib file: " << data);
1587                                 handleFoundFile(data, dep);
1588                         }
1589                 }
1590                 else if (regex_match(token, sub, bibtexError)
1591                          || regex_match(token, sub, bibtexError2)
1592                          || regex_match(token, sub, bibtexError4)
1593                          || regex_match(token, sub, bibtexError5)) {
1594                         retval |= BIBTEX_ERROR;
1595                         string errstr = N_("BibTeX error: ") + token;
1596                         string msg;
1597                         if ((prefixIs(token, "while executing---line")
1598                              || prefixIs(token, "---line ")
1599                              || prefixIs(token, "*Please notify the BibTeX"))
1600                             && !prevtoken.empty()) {
1601                                 errstr = N_("BibTeX error: ") + prevtoken;
1602                                 msg = prevtoken + '\n';
1603                         }
1604                         msg += token;
1605                         terr.insertError(0,
1606                                          from_local8bit(errstr),
1607                                          from_local8bit(msg));
1608                 } else if (regex_match(prevtoken, sub, bibtexError3)) {
1609                         retval |= BIBTEX_ERROR;
1610                         string errstr = N_("BibTeX error: ") + prevtoken;
1611                         string msg = prevtoken + '\n' + token;
1612                         terr.insertError(0,
1613                                          from_local8bit(errstr),
1614                                          from_local8bit(msg));
1615                 } else if (regex_match(token, sub, biberError)) {
1616                         retval |= BIBTEX_ERROR;
1617                         string errstr = N_("Biber error: ") + sub.str(2);
1618                         terr.insertError(0,
1619                                          from_local8bit(errstr),
1620                                          from_local8bit(token));
1621                 }
1622                 prevtoken = token;
1623         }
1624         return retval;
1625 }
1626
1627
1628 int LaTeX::scanIlgFile(TeXErrors & terr)
1629 {
1630         FileName const ilg_file(changeExtension(file.absFileName(), "ilg"));
1631         LYXERR(Debug::OUTFILE, "Scanning ilg file: " << ilg_file);
1632
1633         ifstream ifs(ilg_file.toFilesystemEncoding().c_str());
1634         string token;
1635         int retval = NO_ERRORS;
1636
1637         string prevtoken;
1638         while (getline(ifs, token)) {
1639                 token = rtrim(token, "\r");
1640                 if (prefixIs(token, "!! "))
1641                         prevtoken = token;
1642                 else if (!prevtoken.empty()) {
1643                         retval |= INDEX_ERROR;
1644                         string errstr = N_("Makeindex error: ") + prevtoken;
1645                         string msg = prevtoken + '\n';
1646                         msg += token;
1647                         terr.insertError(0,
1648                                          from_local8bit(errstr),
1649                                          from_local8bit(msg));
1650                         prevtoken.clear();
1651                 } else if (prefixIs(token, "ERROR: ")) {
1652                         retval |= BIBTEX_ERROR;
1653                         string errstr = N_("Xindy error: ") + token.substr(6);
1654                         terr.insertError(0,
1655                                          from_local8bit(errstr),
1656                                          from_local8bit(token));
1657                 }
1658         }
1659         return retval;
1660 }
1661
1662
1663
1664 } // namespace lyx