[igt-dev] [PATCH i-g-t] build: allow to use rst2man from python3

[Petri Latvala]

On Tue, Sep 04, 2018 at 08:35:52PM +0200, Daniel Vetter wrote:
> On Tue, Sep 4, 2018 at 6:05 PM, Lucas De Marchi
> <lucas.demarchi at intel.com> wrote:
> > On Tue, Sep 04, 2018 at 09:00:04AM -0700, Lucas De Marchi wrote:
> >> On Tue, Sep 04, 2018 at 02:10:43PM +0200, Daniel Vetter wrote:
> >> > On Tue, Sep 4, 2018 at 1:56 PM, Jani Nikula <jani.nikula at linux.intel.com> wrote:
> >> > > On Fri, 31 Aug 2018, Lucas De Marchi <lucas.demarchi at intel.com> wrote:
> >> > >> While changing maintainer-tools to allow to use python3 I unsintalled my
> >> > >> python2 tools, which broke IGT build for me. Allow to use either
> >> > >> rst2man-3 or rst2man.
> >> > >>
> >> > >> Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
> >> > >> ---
> >> > >>  configure.ac    | 4 ++--
> >> > >>  man/Makefile.am | 2 +-
> >> > >>  man/meson.build | 4 ++--
> >> > >>  man/rst2man.sh  | 7 ++++---
> >> > >>  4 files changed, 9 insertions(+), 8 deletions(-)
> >> > >>
> >> > >> diff --git a/configure.ac b/configure.ac
> >> > >> index c75ef284..bcd24f04 100644
> >> > >> --- a/configure.ac
> >> > >> +++ b/configure.ac
> >> > >> @@ -52,8 +52,8 @@ enable_gtk_doc=no
> >> > >>  ])
> >> > >>
> >> > >>  # check for rst2man for generating man pages
> >> > >> -AC_CHECK_PROG(RST2MAN, rst2man, yes, no)
> >> > >> -AM_CONDITIONAL(HAVE_RST2MAN, [test "x$RST2MAN" = xyes])
> >> > >> +AC_CHECK_PROGS(RST2MAN, rst2man-3 rst2man, "")
> >> > >> +AM_CONDITIONAL(HAVE_RST2MAN, [test "x$RST2MAN" = x])
> >> > >
> >> > > Is this ugliness what Fedora recommends upstreams all over do to
> >> > > workaround their package management decisions? Yuck.
> >> >
> >> > Use meson (where it's a one-liner). Still meh, but not longer yuck I think :-)
> >> >
> >> > More seriously, do we really care about fixing these kind of things on
> >> > the old automake build system? I'd say just keep it as-is (and only
> >>
> >> Last time I tried[1], Antonio complained about no support for Automake,
> >> Chris ranted that I was "forcing a broken build system that doesn't even
> >> dare to declare itself stable".  Tvrtko said it is important to support
> >> both until autotools is removed. I waited some time to see if any other
> >> maintainer chimed in and then gave up. Now I go with "my life is miserable
> >> and I will update 2 build systems".
> >
> > missed link:
> > https://lists.freedesktop.org/archives/igt-dev/2018-July/004483.html
> 
> I think this is different: Without automake support on this you
> essentially break automake (by making liberal use of a syscall that
> might not exist).
> 
> This here is different: Lack of the python3 alternative won't break
> automake on old/existing systems that want to compile latest igt.
> 
> At least that's my take. tldr;
> - If it's a build system only improvement, ignore automake (as long as
> you don't break it)
> - if it's a change required by new tests/ or lib/ source code, update both.
> 
> Arek/Petri?


Yeah, this matches my expectations as well. Indeed, the memfd_create
change was causing autotools builds to fail _entirely_, an important
distinction.

Only thing keeping us from nuking autotools at this point is
supporting some developers *cough* who still need to build tests with
autotools, and distros who still sometimes want to build tools with
it. CI still (afaik) builds docs with autotools, but that's changing
soon to happen at gitlab directly, at which point we can start by
stripping the support for building docs and man pages with autotools.

Rest follows in a hopefully not-so-distant future. For now the thumb
rules Daniel listed above are the ones to follow.


-- 
Petri Latvala