Re: git: d86a98ea7a48 - main - Uses/electronfix.mk: Rework to not require listing supported electron versions

From: Tomoaki AOKI <junchoon_at_dec.sakura.ne.jp>
Date: Thu, 05 Sep 2024 11:40:57 UTC
On Thu, 5 Sep 2024 09:45:33 +0300
Gleb Popov <arrowd@freebsd.org> wrote:

> On Wed, Sep 4, 2024 at 8:36 PM Tomoaki AOKI <junchoon@dec.sakura.ne.jp> wrote:
> >
> > One thing I can think of is that modifying the offending port (this
> > time, textproc/logseq) to use supported electron (hopefully, the
> > supported one which upstream is currently struggling to work with) and
> > mark it as BROKEN, keeping hooked to build.
> 
> >From my experience, it never works. If the binary distribution was
> built against Electron XY then we need to provide the port with the
> same XY version.
> Sometimes even mismatching patch version (the third number) causes a crash.

My assumption is that:
  *Currently, we have electron 29, 30 and 31 in our ports tree.

  *So regardless which version of electron upstream is currently
   using, and which version upstream is going to use, we should
   choose one from above.

  *Of course, for example, if current version used is 27 and upstream
   is WIP for 28, we should use the nearest 29.

  *As the result, the port should certainly does not work or even
   does not build. So mark it as BROKEN, keeping ports for the time
   upstream catches up (or exceed) with the assumed version.

I assume the above doesn't break INDEX build (not actually tested),
without unhooking (and removing) the port. Because described
dependencies are sane for ports tree (unsane for build, though).
This situation should be just like usual (unwanted) broken ports.


> I never dug into the crash, though.Maybe it is easily fixable, but it
> requires building Electron WITH_DEBUG=yes, which is, well, very
> resource consuming.

As my assumption is keeping the port BROKEN until upstream fixes
the issue, I think it is not needed.

Simply broken (to build, install and/or run) ports certainly affect
its consumers (users, ports depending upon it, directly or indirectly)
but does not affect INDEX builds.

For example, currently at least graphics/nvidia-drm-515-kmod and
graphics/nvidia-drm-61-kmod are broken for stable/14 at commit
183b64dfd84ad7fd504d04f40c90edad809083c1 and later, as the patch
at Bug 279539 [1] is not yet hit the tree.
Already broken for 3 months. But it does not affect for official pkg
buids, just because they are still built against older 14.0.
(Not marked as BROKEN, though.)

[1] https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=279539

Regards.

-- 
Tomoaki AOKI    <junchoon@dec.sakura.ne.jp>