GNOME Bugzilla – Bug 783644
gtk-doc expects source-highlight
Last modified: 2017-08-07 19:54:47 UTC
Created attachment 353548 [details] [review] gtk-doc: disable source-highlight dependency gtk-doc is being built with highlight support enabled, even though it might not be available in the system. Without source-highlight installed, gtkdoc-fixxref fails. No sure whether it is interesting to have source-highlight as sysdep, but I attached a patch that at least disable this depedency. Currently, gnome-software fails with the following traceback: Traceback (most recent call last):
+ Trace 237565
fixxref.Run(options)
FixCrossReferences(options)
FixHTMLFile(options, full_entry)
repl_func, content, flags=re.DOTALL)
return _compile(pattern, flags).sub(repl, string, count)
return HighlightSource(options, m.group(1), m.group(2))
[config.highlight] + shlex.split(highlight_options) + [temp_source_file]).decode('utf-8')
**kwargs).stdout
with Popen(*popenargs, **kwargs) as process:
restore_signals, start_new_session)
raise child_exception_type(errno_num, err_msg)
Review of attachment 353548 [details] [review]: Committed
Syntax highlighting is useful, without knowing it I've compiled several GNOME modules and uploaded tarballs without syntax highlighting in the code examples… For a concrete example, see the code example at: https://developer.gnome.org/gspell/1.5/GspellTextView.html#gspell-text-view-basic-setup it doesn't have syntax highlighting… Thank you very much! Has the patch actually been reviewed before pushing it in jhbuild?
And it is not only about syntax highlighting, normally the functions and types are links.
This is the kind of little thing that makes me sick of the GNOME community. The level is seriously decreasing.
Yes, it was marked as "accepted-commit_now". Considering your comments, I reverted it (commit 2669beac)
To explain a bit more why this was not a good patch to merge, most GNOME maintainers use jhbuild. When doing releases, a GNOME maintainer runs `make distcheck` (with Autotools), which generates the HTML docs for the library. Then the tarball is uploaded, the API documentation is shown on developer.gnome.org and other developers can install the "-doc" package from their Linux distribution to read the documentation in Devhelp. The HTML pages generated during `make distcheck`, run by the maintainer using jhbuild, is what "consumers" of the library will get as documentation. The HTML pages are not (completely) re-generated to show them on developer.gnome.org or to package them in Linux distros.
I'm totally fine with adding a source-highlight sysdep. But obviously we have to disable use of source-highlight until or unless we do so. Sebastien, sorry for the breakage. Please understand we're just trying to keep things building. :) I think this is actually the first time that turning off a feature has affected tarball generation, at least that I remember. It's unfortunate and suggests we need to be especially careful when touching gtk-doc in the future. But the status quo, broken dependency/build, is also not OK.
I see that in the configure.ac of gtk-doc the default is --with-highlight=auto, why is it causing a problem?
I don't know why it is causing a build problem, but you just explained yourself why relying on autodetection is unacceptable. :) Surely source-highlight needs to be a sysdep.
It was maybe just a gtk-doc bug because of the Python porting. Things will be much better with BuildStream.
I don't think it is in gtk-doc, as some other modules were enabling RPM (which as on auto as well) when it was not available. See #783461
Please either add source-highlight to sysdeps (preferred) or else revert this revert so that it's disabled for everyone. I don't want surprise nondeterministic behavior. I agree BuildStream will make this much better. :)
Created attachment 357059 [details] [review] 3.26: gtk-doc depends on source-highlight The attached patch includes source-highlight to sysdeps-3.26 and add it as a dependency of gtk-doc in core-deps-3.26.
Review of attachment 357059 [details] [review]: Thanks!
Review of attachment 357059 [details] [review]: Committed: 9f214f2f7cb58151873f1723734c95e7fdb4bf58
Thank you Rafael. It's normal if I'm a bit pissed off from time to time, don't worry ;)
No problem at all, Sébastien. If it was me in your position, I'd probably get mad as well. Actually, I'm glad you pointed it out, as I really didn't know whether it was a important dependency or not.