3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Alfredo Braunstein
7 * \author Lars Gullik Bjønnes
8 * \author Jean-Marc Lasgouttes
9 * \author Angus Leeming
12 * Full author contact details are available in file CREDITS.
17 #include "bufferlist.h"
24 #include "support/filetools.h"
25 #include "support/convert.h"
26 #include "support/lstrings.h"
27 #include "support/lyxlib.h"
28 #include "support/systemcall.h"
29 #include "support/os.h"
31 #include <boost/filesystem/operations.hpp>
32 #include <boost/regex.hpp>
39 using support::absolutePath;
40 using support::bformat;
41 using support::changeExtension;
42 using support::contains;
43 using support::FileName;
44 using support::findtexfile;
45 using support::getcwd;
46 using support::makeAbsPath;
47 using support::onlyFilename;
48 using support::prefixIs;
49 using support::quoteName;
50 using support::removeExtension;
54 using support::suffixIs;
55 using support::Systemcall;
56 using support::unlink;
59 namespace os = support::os;
60 namespace fs = boost::filesystem;
66 #ifndef CXX_GLOBAL_CSTD
77 // TODO: in no particular order
78 // - get rid of the call to
79 // BufferList::updateIncludedTeXfiles, this should either
80 // be done before calling LaTeX::funcs or in a completely
82 // - the makeindex style files should be taken care of with
83 // the dependency mechanism.
84 // - makeindex commandline options should be supported
85 // - somewhere support viewing of bibtex and makeindex log files.
86 // - we should perhaps also scan the bibtex log file
90 docstring runMessage(unsigned int count)
92 return bformat(_("Waiting for LaTeX run number %1$d"), count);
101 void TeXErrors::insertError(int line, docstring const & error_desc,
102 docstring const & error_text)
104 Error newerr(line, error_desc, error_text);
105 errors.push_back(newerr);
109 bool operator==(Aux_Info const & a, Aux_Info const & o)
111 return a.aux_file == o.aux_file &&
112 a.citations == o.citations &&
113 a.databases == o.databases &&
114 a.styles == o.styles;
118 bool operator!=(Aux_Info const & a, Aux_Info const & o)
128 LaTeX::LaTeX(string const & latex, OutputParams const & rp,
130 : cmd(latex), file(f), runparams(rp)
133 if (prefixIs(cmd, "pdf")) { // Do we use pdflatex ?
134 depfile = FileName(file.absFilename() + ".dep-pdf");
135 output_file = FileName(changeExtension(file.absFilename(), ".pdf"));
137 depfile = FileName(file.absFilename() + ".dep");
138 output_file = FileName(changeExtension(file.absFilename(), ".dvi"));
143 void LaTeX::deleteFilesOnError() const
145 // currently just a dummy function.
147 // What files do we have to delete?
149 // This will at least make latex do all the runs
152 // but the reason for the error might be in a generated file...
155 FileName const bbl(changeExtension(file.absFilename(), ".bbl"));
159 FileName const ind(changeExtension(file.absFilename(), ".ind"));
163 FileName const nls(changeExtension(file.absFilename(), ".nls"));
166 // Also remove the aux file
167 FileName const aux(changeExtension(file.absFilename(), ".aux"));
172 int LaTeX::run(TeXErrors & terr)
173 // We know that this function will only be run if the lyx buffer
174 // has been changed. We also know that a newly written .tex file
175 // is always different from the previous one because of the date
176 // in it. However it seems safe to run latex (at least) on time each
177 // time the .tex file changes.
179 int scanres = NO_ERRORS;
180 unsigned int count = 0; // number of times run
181 num_errors = 0; // just to make sure.
182 unsigned int const MAX_RUN = 6;
183 DepTable head; // empty head
184 bool rerun = false; // rerun requested
186 // The class LaTeX does not know the temp path.
187 theBufferList().updateIncludedTeXfiles(getcwd().absFilename(), runparams);
189 // Never write the depfile if an error was encountered.
192 // first check if the file dependencies exist:
193 // ->If it does exist
194 // check if any of the files mentioned in it have
195 // changed (done using a checksum).
197 // run latex once and
198 // remake the dependency file
199 // -> if not changed:
200 // just return there is nothing to do for us.
201 // ->if it doesn't exist
203 // run latex once (we need to run latex once anyway) and
204 // remake the dependency file.
207 bool had_depfile = fs::exists(depfile.toFilesystemEncoding());
208 bool run_bibtex = false;
209 FileName const aux_file(changeExtension(file.absFilename(), "aux"));
212 lyxerr[Debug::DEPEND] << "Dependency file exists" << endl;
213 // Read the dep file:
214 had_depfile = head.read(depfile);
218 // Update the checksums
220 // Can't just check if anything has changed because it might have aborted
221 // on error last time... in which cas we need to re-run latex
222 // and collect the error messages (even if they are the same).
223 if (!fs::exists(output_file.toFilesystemEncoding())) {
224 lyxerr[Debug::DEPEND]
225 << "re-running LaTeX because output file doesn't exist." << endl;
226 } else if (!head.sumchange()) {
227 lyxerr[Debug::DEPEND] << "return no_change" << endl;
230 lyxerr[Debug::DEPEND]
231 << "Dependency file has changed" << endl;
234 if (head.extchanged(".bib") || head.extchanged(".bst"))
237 lyxerr[Debug::DEPEND]
238 << "Dependency file does not exist, or has wrong format" << endl;
240 /// We scan the aux file even when had_depfile = false,
241 /// because we can run pdflatex on the file after running latex on it,
242 /// in which case we will not need to run bibtex again.
243 vector<Aux_Info> bibtex_info_old;
245 bibtex_info_old = scanAuxFiles(aux_file);
248 lyxerr[Debug::LATEX] << "Run #" << count << endl;
249 message(runMessage(count));
252 scanres = scanLogFile(terr);
253 if (scanres & ERROR_RERUN) {
254 lyxerr[Debug::LATEX] << "Rerunning LaTeX" << endl;
256 scanres = scanLogFile(terr);
259 if (scanres & ERRORS) {
260 deleteFilesOnError();
261 return scanres; // return on error
264 vector<Aux_Info> const bibtex_info = scanAuxFiles(aux_file);
265 if (!run_bibtex && bibtex_info_old != bibtex_info)
268 // update the dependencies.
269 deplog(head); // reads the latex log
273 // At this point we must run external programs if needed.
274 // makeindex will be run if a .idx file changed or was generated.
275 // And if there were undefined citations or changes in references
276 // the .aux file is checked for signs of bibtex. Bibtex is then run
279 // memoir (at least) writes an empty *idx file in the first place.
280 // A second latex run is needed.
281 FileName const idxfile(changeExtension(file.absFilename(), ".idx"));
282 rerun = fs::exists(idxfile.toFilesystemEncoding()) &&
283 fs::is_empty(idxfile.toFilesystemEncoding());
286 if (head.haschanged(idxfile)) {
288 lyxerr[Debug::LATEX] << "Running MakeIndex." << endl;
289 message(_("Running MakeIndex."));
290 // onlyFilename() is needed for cygwin
291 rerun |= runMakeIndex(onlyFilename(idxfile.absFilename()), runparams);
293 if (head.haschanged(FileName(changeExtension(file.absFilename(), ".nlo")))) {
294 lyxerr[Debug::LATEX] << "Running MakeIndex for nomencl." << endl;
295 message(_("Running MakeIndex for nomencl."));
296 // onlyFilename() is needed for cygwin
297 string const nomenclstr = " -s nomencl.ist -o " + onlyFilename(changeExtension(file.toFilesystemEncoding(), ".nls"));
298 rerun |= runMakeIndex(onlyFilename(changeExtension(file.absFilename(), ".nlo")), runparams, nomenclstr);
302 // if (scanres & UNDEF_CIT || scanres & RERUN || run_bibtex)
303 if (scanres & UNDEF_CIT || run_bibtex) {
304 // Here we must scan the .aux file and look for
305 // "\bibdata" and/or "\bibstyle". If one of those
306 // tags is found -> run bibtex and set rerun = true;
308 lyxerr[Debug::LATEX] << "Running BibTeX." << endl;
309 message(_("Running BibTeX."));
310 updateBibtexDependencies(head, bibtex_info);
311 rerun |= runBibTeX(bibtex_info);
312 } else if (!had_depfile) {
313 /// If we run pdflatex on the file after running latex on it,
314 /// then we do not need to run bibtex, but we do need to
315 /// insert the .bib and .bst files into the .dep-pdf file.
316 updateBibtexDependencies(head, bibtex_info);
320 // we know on this point that latex has been run once (or we just
321 // returned) and the question now is to decide if we need to run
322 // it any more. This is done by asking if any of the files in the
323 // dependency file has changed. (remember that the checksum for
324 // a given file is reported to have changed if it just was created)
325 // -> if changed or rerun == true:
326 // run latex once more and
327 // update the dependency structure
328 // -> if not changed:
329 // we does nothing at this point
331 if (rerun || head.sumchange()) {
334 lyxerr[Debug::DEPEND]
335 << "Dep. file has changed or rerun requested" << endl;
337 << "Run #" << count << endl;
338 message(runMessage(count));
340 scanres = scanLogFile(terr);
341 if (scanres & ERRORS) {
342 deleteFilesOnError();
343 return scanres; // return on error
346 // update the depedencies
347 deplog(head); // reads the latex log
350 lyxerr[Debug::DEPEND] << "Dep. file has NOT changed" << endl;
354 // The inclusion of files generated by external programs like
355 // makeindex or bibtex might have done changes to pagenumbering,
356 // etc. And because of this we must run the external programs
357 // again to make sure everything is redone correctly.
358 // Also there should be no need to run the external programs any
361 // run makeindex if the <file>.idx has changed or was generated.
362 if (head.haschanged(FileName(changeExtension(file.absFilename(), ".idx")))) {
364 lyxerr[Debug::LATEX] << "Running MakeIndex." << endl;
365 message(_("Running MakeIndex."));
366 // onlyFilename() is needed for cygwin
367 rerun = runMakeIndex(onlyFilename(changeExtension(file.absFilename(), ".idx")), runparams);
370 // I am not pretty sure if need this twice.
371 if (head.haschanged(FileName(changeExtension(file.absFilename(), ".nlo")))) {
372 lyxerr[Debug::LATEX] << "Running MakeIndex for nomencl." << endl;
373 message(_("Running MakeIndex for nomencl."));
374 // onlyFilename() is needed for cygwin
375 string nomenclstr = " -s nomencl.ist -o " + onlyFilename(changeExtension(file.toFilesystemEncoding(), ".nls"));
376 rerun |= runMakeIndex(onlyFilename(changeExtension(file.absFilename(), ".nlo")), runparams, nomenclstr);
380 // we will only run latex more if the log file asks for it.
381 // or if the sumchange() is true.
382 // -> rerun asked for:
384 // remake the dependency file
385 // goto 2 or return if max runs are reached.
386 // -> rerun not asked for:
387 // just return (fall out of bottom of func)
389 while ((head.sumchange() || rerun || (scanres & RERUN))
390 && count < MAX_RUN) {
391 // Yes rerun until message goes away, or until
392 // MAX_RUNS are reached.
395 lyxerr[Debug::LATEX] << "Run #" << count << endl;
396 message(runMessage(count));
398 scanres = scanLogFile(terr);
399 if (scanres & ERRORS) {
400 deleteFilesOnError();
401 return scanres; // return on error
408 // Write the dependencies to file.
410 lyxerr[Debug::LATEX] << "Done." << endl;
415 int LaTeX::startscript()
417 // onlyFilename() is needed for cygwin
418 string tmp = cmd + ' ' + quoteName(onlyFilename(file.toFilesystemEncoding())) + " > " + os::nulldev();
420 return one.startscript(Systemcall::Wait, tmp);
424 bool LaTeX::runMakeIndex(string const & f, OutputParams const & runparams,
425 string const & params)
428 << "idx file has been made, running makeindex on file "
430 string tmp = lyxrc.index_command + ' ';
432 tmp = subst(tmp, "$$lang", runparams.document_language);
436 one.startscript(Systemcall::Wait, tmp);
441 vector<Aux_Info> const
442 LaTeX::scanAuxFiles(FileName const & file)
444 vector<Aux_Info> result;
446 result.push_back(scanAuxFile(file));
448 string const basename = removeExtension(file.absFilename());
449 for (int i = 1; i < 1000; ++i) {
450 FileName const file2(basename
451 + '.' + convert<string>(i)
453 if (!fs::exists(file2.toFilesystemEncoding()))
455 result.push_back(scanAuxFile(file2));
461 Aux_Info const LaTeX::scanAuxFile(FileName const & file)
464 result.aux_file = file;
465 scanAuxFile(file, result);
470 void LaTeX::scanAuxFile(FileName const & file, Aux_Info & aux_info)
472 lyxerr[Debug::LATEX] << "Scanning aux file: " << file << endl;
474 ifstream ifs(file.toFilesystemEncoding().c_str());
476 static regex const reg1("\\\\citation\\{([^}]+)\\}");
477 static regex const reg2("\\\\bibdata\\{([^}]+)\\}");
478 static regex const reg3("\\\\bibstyle\\{([^}]+)\\}");
479 static regex const reg4("\\\\@input\\{([^}]+)\\}");
481 while (getline(ifs, token)) {
482 token = rtrim(token, "\r");
484 // FIXME UNICODE: We assume that citation keys and filenames
485 // in the aux file are in the file system encoding.
486 token = to_utf8(from_filesystem8bit(token));
487 if (regex_match(token, sub, reg1)) {
488 string data = sub.str(1);
489 while (!data.empty()) {
491 data = split(data, citation, ',');
492 lyxerr[Debug::LATEX] << "Citation: "
494 aux_info.citations.insert(citation);
496 } else if (regex_match(token, sub, reg2)) {
497 string data = sub.str(1);
498 // data is now all the bib files separated by ','
499 // get them one by one and pass them to the helper
500 while (!data.empty()) {
502 data = split(data, database, ',');
503 database = changeExtension(database, "bib");
504 lyxerr[Debug::LATEX] << "BibTeX database: `"
505 << database << '\'' << endl;
506 aux_info.databases.insert(database);
508 } else if (regex_match(token, sub, reg3)) {
509 string style = sub.str(1);
510 // token is now the style file
511 // pass it to the helper
512 style = changeExtension(style, "bst");
513 lyxerr[Debug::LATEX] << "BibTeX style: `"
514 << style << '\'' << endl;
515 aux_info.styles.insert(style);
516 } else if (regex_match(token, sub, reg4)) {
517 string const file2 = sub.str(1);
518 scanAuxFile(makeAbsPath(file2), aux_info);
524 void LaTeX::updateBibtexDependencies(DepTable & dep,
525 vector<Aux_Info> const & bibtex_info)
527 // Since a run of Bibtex mandates more latex runs it is ok to
528 // remove all ".bib" and ".bst" files.
529 dep.remove_files_with_extension(".bib");
530 dep.remove_files_with_extension(".bst");
531 //string aux = OnlyFilename(ChangeExtension(file, ".aux"));
533 for (vector<Aux_Info>::const_iterator it = bibtex_info.begin();
534 it != bibtex_info.end(); ++it) {
535 for (set<string>::const_iterator it2 = it->databases.begin();
536 it2 != it->databases.end(); ++it2) {
537 FileName const file = findtexfile(*it2, "bib");
539 dep.insert(file, true);
542 for (set<string>::const_iterator it2 = it->styles.begin();
543 it2 != it->styles.end(); ++it2) {
544 FileName const file = findtexfile(*it2, "bst");
546 dep.insert(file, true);
552 bool LaTeX::runBibTeX(vector<Aux_Info> const & bibtex_info)
555 for (vector<Aux_Info>::const_iterator it = bibtex_info.begin();
556 it != bibtex_info.end(); ++it) {
557 if (it->databases.empty())
561 string tmp = lyxrc.bibtex_command + " ";
562 // onlyFilename() is needed for cygwin
563 tmp += quoteName(onlyFilename(removeExtension(it->aux_file.absFilename())));
565 one.startscript(Systemcall::Wait, tmp);
567 // Return whether bibtex was run
572 int LaTeX::scanLogFile(TeXErrors & terr)
576 int retval = NO_ERRORS;
577 string tmp = onlyFilename(changeExtension(file.absFilename(), ".log"));
578 lyxerr[Debug::LATEX] << "Log file: " << tmp << endl;
579 FileName const fn = FileName(makeAbsPath(tmp));
580 ifstream ifs(fn.toFilesystemEncoding().c_str());
583 while (getline(ifs, token)) {
584 // MikTeX sometimes inserts \0 in the log file. They can't be
585 // removed directly with the existing string utility
586 // functions, so convert them first to \r, and remove all
587 // \r's afterwards, since we need to remove them anyway.
588 token = subst(token, '\0', '\r');
589 token = subst(token, "\r", "");
591 lyxerr[Debug::LATEX] << "Log line: " << token << endl;
596 if (prefixIs(token, "LaTeX Warning:") ||
597 prefixIs(token, "! pdfTeX warning")) {
598 // Here shall we handle different
600 retval |= LATEX_WARNING;
601 lyxerr[Debug::LATEX] << "LaTeX Warning." << endl;
602 if (contains(token, "Rerun to get cross-references")) {
605 << "We should rerun." << endl;
606 } else if (contains(token, "Citation")
607 && contains(token, "on page")
608 && contains(token, "undefined")) {
611 } else if (prefixIs(token, "Package")) {
613 retval |= PACKAGE_WARNING;
614 if (contains(token, "natbib Warning:")) {
616 if (contains(token, "Citation")
617 && contains(token, "on page")
618 && contains(token, "undefined")) {
621 } else if (contains(token, "run BibTeX")) {
623 } else if (contains(token, "Rerun LaTeX") ||
624 contains(token, "Rerun to get")) {
625 // at least longtable.sty and bibtopic.sty
628 << "We should rerun." << endl;
631 } else if (token[0] == '(') {
632 if (contains(token, "Rerun LaTeX") ||
633 contains(token, "Rerun to get")) {
636 << "We should rerun." << endl;
639 } else if (prefixIs(token, "! ")) {
640 // Ok, we have something that looks like a TeX Error
641 // but what do we really have.
643 // Just get the error description:
644 string desc(token, 2);
645 if (contains(token, "LaTeX Error:"))
646 retval |= LATEX_ERROR;
651 if (!getline(ifs, tmp))
655 } while (!prefixIs(tmp, "l."));
656 if (prefixIs(tmp, "l.")) {
657 // we have a latex error
659 if (contains(desc, "Package babel Error: You haven't defined the language"))
660 retval |= ERROR_RERUN;
661 // get the line number:
663 sscanf(tmp.c_str(), "l.%d", &line);
664 // get the rest of the message:
665 string errstr(tmp, tmp.find(' '));
668 while (!contains(errstr, "l.")
670 && !prefixIs(tmp, "! ")
671 && !contains(tmp, "(job aborted")) {
677 << "line: " << line << '\n'
678 << "Desc: " << desc << '\n'
679 << "Text: " << errstr << endl;
680 if (line == last_line)
686 if (line_count <= 5) {
688 // We have no idea what the encoding of
690 // It seems that the output from the
691 // latex compiler itself is pure ASCII,
692 // but it can include bits from the
693 // document, so whatever encoding we
694 // assume here it can be wrong.
695 terr.insertError(line, from_local8bit(desc), from_local8bit(errstr));
700 // information messages, TeX warnings and other
701 // warnings we have not caught earlier.
702 if (prefixIs(token, "Overfull ")) {
703 retval |= TEX_WARNING;
704 } else if (prefixIs(token, "Underfull ")) {
705 retval |= TEX_WARNING;
706 } else if (contains(token, "Rerun to get citations")) {
707 // Natbib seems to use this.
709 } else if (contains(token, "No pages of output")) {
710 // A dvi file was not created
712 } else if (contains(token, "That makes 100 errors")) {
713 // More than 100 errors were reprted
714 retval |= TOO_MANY_ERRORS;
718 lyxerr[Debug::LATEX] << "Log line: " << token << endl;
725 void handleFoundFile(string const & ff, DepTable & head)
727 // convert from native os path to unix path
728 string const foundfile = os::internal_path(trim(ff));
730 lyxerr[Debug::DEPEND] << "Found file: " << foundfile << endl;
732 // Ok now we found a file.
733 // Now we should make sure that this is a file that we can
734 // access through the normal paths.
735 // We will not try any fancy search methods to
738 // (1) foundfile is an
739 // absolute path and should
741 if (absolutePath(foundfile)) {
742 lyxerr[Debug::DEPEND] << "AbsolutePath file: "
743 << foundfile << endl;
744 // On initial insert we want to do the update at once
745 // since this file cannot be a file generated by
747 FileName const absname(foundfile);
748 if (fs::exists(absname.toFilesystemEncoding()) &&
749 !fs::is_directory(absname.toFilesystemEncoding()))
750 head.insert(absname, true);
755 string const onlyfile = onlyFilename(foundfile);
756 FileName const absname(makeAbsPath(onlyfile));
758 // (2) foundfile is in the tmpdir
759 // insert it into head
760 if (fs::exists(absname.toFilesystemEncoding())) {
761 static regex unwanted("^.*\\.(aux|log|dvi|bbl|ind|glo)$");
762 if (regex_match(onlyfile, unwanted)) {
763 lyxerr[Debug::DEPEND]
766 << " in the dep file"
768 } else if (suffixIs(onlyfile, ".tex")) {
769 // This is a tex file generated by LyX
770 // and latex is not likely to change this
772 lyxerr[Debug::DEPEND]
773 << "Tmpdir TeX file: "
776 head.insert(absname, true);
778 lyxerr[Debug::DEPEND]
782 head.insert(absname);
785 lyxerr[Debug::DEPEND]
786 << "Not a file or we are unable to find it."
793 void LaTeX::deplog(DepTable & head)
795 // This function reads the LaTeX log file end extracts all the external
796 // files used by the LaTeX run. The files are then entered into the
799 string const logfile = onlyFilename(changeExtension(file.absFilename(), ".log"));
801 static regex reg1(".*\\([^)]+.*");
802 static regex reg2("File: ([^ ]+).*");
803 static regex reg3("No file ([^ ]+)\\..*");
804 static regex reg4("\\\\openout[0-9]+.*=.*`([^ ]+)'\\..*");
805 // If an index should be created, MikTex does not write a line like
806 // \openout# = 'sample.idx'.
807 // but instead only a line like this into the log:
808 // Writing index file sample.idx
809 static regex reg5("Writing index file ([^ ]+).*");
810 static regex regnomencl("Writing nomenclature file ([^ ]+).*");
811 // If a toc should be created, MikTex does not write a line like
812 // \openout# = `sample.toc'.
813 // but only a line like this into the log:
815 // This line is also written by tetex.
816 // This line is not present if no toc should be created.
817 static regex miktexTocReg("\\\\tf@toc=\\\\write.*");
819 FileName const fn(makeAbsPath(logfile));
820 ifstream ifs(fn.toFilesystemEncoding().c_str());
822 // Ok, the scanning of files here is not sufficient.
823 // Sometimes files are named by "File: xxx" only
824 // So I think we should use some regexps to find files instead.
825 // "(\([^ ]+\)" should match the "(file " variant, note
826 // that we can have several of these on one line.
827 // "File: \([^ ]+\)" should match the "File: file" variant
831 // MikTeX sometimes inserts \0 in the log file. They can't be
832 // removed directly with the existing string utility
833 // functions, so convert them first to \r, and remove all
834 // \r's afterwards, since we need to remove them anyway.
835 token = subst(token, '\0', '\r');
836 token = subst(token, "\r", "");
842 // FIXME UNICODE: We assume that the file names in the log
843 // file are in the file system encoding.
844 token = to_utf8(from_filesystem8bit(token));
846 if (regex_match(token, sub, reg1)) {
847 static regex reg1_1("\\(([^()]+)");
849 string::const_iterator first = token.begin();
850 string::const_iterator end = token.end();
852 while (regex_search(first, end, what, reg1_1)) {
853 first = what[0].second;
854 handleFoundFile(what.str(1), head);
856 } else if (regex_match(token, sub, reg2))
857 handleFoundFile(sub.str(1), head);
858 else if (regex_match(token, sub, reg3))
859 handleFoundFile(sub.str(1), head);
860 else if (regex_match(token, sub, reg4))
861 handleFoundFile(sub.str(1), head);
862 else if (regex_match(token, sub, reg5))
863 handleFoundFile(sub.str(1), head);
864 else if (regex_match(token, sub, regnomencl))
865 handleFoundFile(sub.str(1), head);
866 else if (regex_match(token, sub, miktexTocReg))
867 handleFoundFile(onlyFilename(changeExtension(file.absFilename(), ".toc")), head);
870 // Make sure that the main .tex file is in the dependancy file.
871 head.insert(file, true);