Am Dienstag, 17. Februar 2004 18:09 schrieb Jean-Marc Lasgouttes: > I am sorry I have not been more specific at the time.
Never mind, fortunately it does not change anything fundamental, at least I hope so ;-) > Georg> I thought that the "run in original dir" flag was used for html > Georg> converters? I guess that if this flag is honoured correctly > Georg> html export should also work with a temp dir, putting the > Georg> generated files deliberately not in the tmp dir. > > It seems that the flag has been designed for html, but someone > (Herbert?) at some time decided to remove the flag. Does this means > that it had bad side-effects? I do not know. We should also check that > originaldir still works. It is still there, but I believe not fully functional. > Basically, there are two approaches: > > - put everything in the master buffer temp dir > > - put each file in its own document temp dir, but use an absolute > file name to reference it (this is what insetgraphics does, I think) > > What you say is that the first approach avoids problems with dvi, > right? Yes, because we can use relative filenames. This makes it easy to copy the dvi + graphics after the latex run. If the dvi would contain absolute filenames as it is now, one would have to edit the dvi as Angus suggested. > Another thing that you may want to consider is the get rid of the > [EMAIL PROTECTED] hack and directly modify TEXINPUTS instead. Any file found > through [EMAIL PROTECTED] will show up as an absolute path in the dvi, whih > _may_ be a problem. Files found through TEXINPUTS appear as relative > paths. I believe that this hack is not needed anymore, at least not for the graphics inset. But Helges example shows that it could be useful for other reasons. > I have been against modifying TEXINPUTS for a long time because it is > difficult to do in a portable way. However, I begin to wonder whether > it would be a better alternative. What is difficult: Setting the environment variable, or determine its value? I guess the latter should be doable without too much pain with the existing path conversion routines? > OK, now that you have explained what the complete plan is, I agree too > that we can try to go this way. > > Do you want me to apply your patch now? Not yet, I first want to check wether the html stuff works as expected. I'll send an updated version then, including the warning message that Angus suggested. Georg