On 21 August 2015 at 12:42, Gavin Smith <[email protected]> wrote: >> Your email confirms what I thought, my problem is that my log file does >> not contain any \openout traces while yours have them. Please note that >> your diff did not work properly probably because my log uses CRLF EOL's >> and your uses LF. To make diff work properly, please try: >> >> diff -ab metafont-for-beginners.log{,-VINCENT} > > Isn't metafont-for-beginners.log just that log of the final (e)TeX > run, so if texi2dvi runs it twice, we won't see what happened from the > first time that caused it to be read the second time? > >> So it seems that the difference comes from this that MikTeX does not >> make these \openout traces. After the first run you have: >> >> /usr/local/bin/texi2dvi: Original xref files = >> /usr/local/bin/texi2dvi: New xref files = metafont-for-beginners.aux >> metafont-for-beginners.toc >> >> while in my case, I have: >> >> /local/bin/texi2dvi: Original xref files = >> /local/bin/texi2dvi: New xref files = >> >> ie. New xref file list is empty. >> >> One more thing that strikes me is the following: your log file contains >> the \openout traces, but these traces do not appear in the output of >> texi2dvi, so there are traced as information (going to the log file but >> not to the standard output). So something seems to prevent MikTeX from >> logging the information, maybe the program is stopped because this >> stream is flushed. >> >> One more thing I cannot understand: why don't the other output files >> (index files .ky, .cp, etc...) go tot he New xref files list in your >> case ? > > From my reading of the source of texi2dvi, it looks in the filesystem > to see what files were produced; it doesn't look at what \openout's > were done.
Disregard that, you are right. So it does appear that the problem is that the \openout lines are missing from the log file. >> - why index files are overlooked by texi2dvi, ie not in the xref file >> list? Maybe because they are empty, and so don't pass xref_file_p?
