On Sun, Nov 22, 2020 at 09:02:17AM +0000, Gavin Smith wrote: > On Sun, Nov 22, 2020 at 12:46:35AM +0100, Patrice Dumas wrote: > > I think that > > the point is to copy the image file from where it is found to the > > destination directory. > > Could this break the build systems of any existing manuals that use images? > Are there any manuals that we could check?
It could be possible to reach GNU maintainers for GNU manuals if needed and also to post on gnu-system-discuss (though I am not 100% sure who reads this list). > Also, what if the image files are out-of-date? I think that we should not care about that case. As long as the file exists, even if, for instance it is a broken link, we do nothing. > I looked at the gendocs.sh script from gnulib, which would be one of > the most usual ways for people to generate HTML manuals that they > will upload to a website. It does take care of copying images, > copying them after running texi2any. This should be harmless if it > does it twice, just potentially slow if a manual has many images, but > gendocs.sh could potentially be altered to check if the files already > exist. Indeed. > It could be a benefit if texi2any does it instead of gendocs.sh, as there > is this comment in gendocs.sh, at the copy_images function: > > # copy_images OUTDIR HTML-FILE... > # ------------------------------- > # Copy all the images needed by the HTML-FILEs into OUTDIR. > # Look for them in . and the -I directories; this is simpler than what > # makeinfo supports with -I, but hopefully it will suffice. I think that this is indeed better if texi2any does it instead of an emulation in gendocs.sh, less risks for confusion. -- Pat