Re: [libvirt PATCH] meson: Ask rst2html to strip comments

On Mon, Jun 07, 2021 at 04:52:41PM +0200, Pavel Hrdina wrote: > On Mon, Jun 07, 2021 at 04:39:15PM +0200, Michal Prívozník wrote: >> On 6/4/21 2:31 PM, Andrea Bolognani wrote: >>> They can be problematic: in particular, the rst files generated >>> by keycodemapdb's keymap-gen contain things like >>> >>> To re-generate, run: >>> keymap-gen --lang=rst --title=virkeycode-osx [...] >>> >>> which result in xsltproc later choking with >>> >>> [1/12] Generating virkeyname-osx.html with a meson_exe.py custom command >>> FAILED: docs/manpages/virkeyname-osx.html >>> /usr/bin/meson --internal exe --capture docs/manpages/virkeyname-osx.html \ >>> /usr/bin/xsltproc [...] --nonet ../docs/subsite.xsl docs/manpages/virkeyname-osx.html.in >>> docs/manpages/virkeyname-osx.html.in:17: parser error : Double hyphen within comment: >>> keymap-gen --lang=rst --title=virkeyname-osx [...] >> >> I don't see this error. Do I need some very fresh version of something? >> >> $ rst2html.py --version >> rst2html.py (Docutils 0.16 [release], Python 3.9.4, on linux) > > You need the incorrect rst2html tool, see this explanation for example: > > https://gitlab.com/libvirt/libvirt/-/issues/139#note_528736524 > > I'm not so sure about this patch. I think we should instead try to > detect what rst2html is used and error out if it is not the one provided > by docutils project. I've looked detecting the wrong rst2html "binary" before, but I didn't see any attractive ways todo it.

I wonder if the docutils python API is considered stable ? If it is, then we could skip calling the system rst2html5 entirely, and just copy its contents into scripts/rst2html5. Then all we need to check is that docutils python API is installed, and we'll know we run the right thing.