On Fri, Jun 05, 2020 at 03:02:05AM +0800, Xi Ruoyao via blfs-dev wrote:
> On 2020-06-04 12:38 +0100, Ken Moffat via blfs-dev wrote:
> > (starting a new thread so that people will notice this :)
> >
> > Back in December I built an experimental system to try doing without
> > python2. At that time it was less than useful (e.g. no graphical
> > browsers, and I think I had to build suid xorg because I couldn't
> > build mozilla's JS). But I kept the system for testing.
>
> I've been use such an "experimental" system for a while. The main blockers
> for
> me are js{60,68} and asciidoc. For js68 I made a giant patch. For js60
> (which
> is only required by polkit now) I made a small polkit patch to use js68
> instead
> (which is now a MR in polkit repo). For asciidoc I used asciidoc-9.0.0rc{1,2}
> (good news: 9.0.0 has been released).
>
Yes, I remember your js68 patch. For me (sysv) polkit is one of the
flies in the ointment. But others are qtwebengine (probably) and
gimp (until 3.0). My experimental system has only been retained to
see if I could build firefox without py2.
But it is now so old (the clang and gcc versions) that I'll probably
need to trash it soon.
> >
> > Please note that for 78.0b2 I'm using the default compiler (clang) -
> > on another machine where I do my firefox beta testing the build with
> > gcc (9.3.0) failed (Error but without any obvious detailed error
> > message to explain what happened - but of course lots of output which
> > happend to mention error or :Error in compiler warnings) so for the
> > moment I'm using clang.
>
> It's either a Firefox bug or a GCC bug. Normally I would make a bug report
> but
> I'm too lazy to build Firefox (I would need 4 additional packages for it:
> GTK+-
> 2, cbindgen, nodejs, and libevent).
>
Yeah, the problem with all the rust code is that it becomes
impossible to determine where/why it fails. But I've seen this
before, so I'm hopeful the problem will again disappear before the
release. And I need to build more gcc-10 systems to look at this.
Usually I build python2 as a matter of course, the problem with that
is that it gets tedious to switch out all the py2 programs and
libraries. I suppsoe I'll need to add some scripts to
disable/enable py2 on a completed system.
> One Mozilla developer once made a lecture in our university and said "do not
> try
> to build Firefox without a powerful workstation" :).
I bet that was years ago! The build has got slower, and
continues to get slower. And mozilla devs say that current
llvm makes rust compiles slower. But against that, for firefox
development they run all the tests and do some LTO and PGO to
produce their binaries which of course adds a lot more overhead to
the build. I've got a maximum of 8 cores (4 + hyperthreading) and
watching 'top' I can see that the build has several bottlenecks
where only one core is being used.
Still, one step nearer ditching py2 has to be good.
ĸen
--
+++ OUT OF CHEESE ERROR. REDO FROM START +++
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
