[gentoo-dev] What happened to gcc-12.3.0?

[gentoo-dev] What happened to gcc-12.3.0?

[Joshua Kinard]

Noticing that the ebuild for gcc-12.3.0 got dropped with little explanation.  It is the upstream stable 
release.  I am eyeballing #906310 as what may have triggered the drop, but I find it a bit of a stretch that 
an upstream stable release got dropped over a single, optional package that has a history of quirky behavior 
(FWIW, I never had luck with ccache, especially on MIPS).

I know we have the pYYYMMDD ebuilds, but I've been keeping my mips chroots on only the upstream/stable 
releases to minimize the number of times I have to rebuild the compiler, since we make frequent pYYYYMMDD 
releases (once a month, on average, judging by the datestamps), and most of the patches in the pYYYYMMDD 
series don't help or hinder MIPS.

Under qemu, it takes about 4 hours to build the single-ABI variant of gcc and 7 hours for the multilib 
variant.  So I avoid rebuilding the compiler as much as possible, as with six chroots, that's virtually an 
entire day across all six just for gcc, minus distractions (seriously, the build times on gcc are getting 
waaaaaaay out of hand, regardless of arch).

Options?  I mean, if anyone knows magic to make gcc build faster, I am all ears, but ever since the switch to 
C++, the time needed for it to build itself has just been absolutely horrendous.  And it gets worse with each 
new release, for some reason.

-- 
Joshua Kinard
Gentoo/MIPS
kumba@gentoo.org
rsa6144/5C63F4E3F5C6C943 2015-04-27
177C 1972 1FB8 F254 BAD0 3E72 5C63 F4E3 F5C6 C943

"The past tempts us, the present confuses us, the future frightens us.  And our lives slip away, moment by 
moment, lost in that vast, terrible in-between."

         --Emperor Turhan, Centauri Republic

Re: [gentoo-dev] What happened to gcc-12.3.0?

[Matt Turner]

On Thu, Jun 15, 2023 at 12:02 AM Joshua Kinard <kumba@gentoo.org> wrote:
 Options?  I mean, if anyone knows magic to make gcc build faster, I
am all ears, but ever since the switch to
> C++, the time needed for it to build itself has just been absolutely horrendous.  And it gets worse with each
> new release, for some reason.

EXTRA_ECONF=--disable-bootstrap

See https://bugs.gentoo.org/705406#c1

Re: [gentoo-dev] What happened to gcc-12.3.0?

[Joshua Kinard]

On 6/15/2023 01:04, Matt Turner wrote:
> On Thu, Jun 15, 2023 at 12:02 AM Joshua Kinard <kumba@gentoo.org> wrote:
>   Options?  I mean, if anyone knows magic to make gcc build faster, I
> am all ears, but ever since the switch to
>> C++, the time needed for it to build itself has just been absolutely horrendous.  And it gets worse with each
>> new release, for some reason.
> 
> EXTRA_ECONF=--disable-bootstrap
> 
> See https://bugs.gentoo.org/705406#c1

Thanks, I'll give this a go and see what the build time differences are like.  I always thought the extra 
build phase was unnecessary for typical builds, but I never recalled gcc having this switch before.  I don't 
exactly play around in its internals much these days, so I probably missed its addition.

Any thought been given to making it into a feature controlled by a USE flag?  E.g., 'full-bootstrap' or 
something, defaulting to on, to build all three stages, and users can turn it off to skip the comparison stage?

-- 
Joshua Kinard
Gentoo/MIPS
kumba@gentoo.org
rsa6144/5C63F4E3F5C6C943 2015-04-27
177C 1972 1FB8 F254 BAD0 3E72 5C63 F4E3 F5C6 C943

"The past tempts us, the present confuses us, the future frightens us.  And our lives slip away, moment by 
moment, lost in that vast, terrible in-between."

         --Emperor Turhan, Centauri Republic

Re: [gentoo-dev] What happened to gcc-12.3.0?

[Sam James]

[-- Attachment #1: Type: text/plain, Size: 1265 bytes --]


Joshua Kinard <kumba@gentoo.org> writes:

> Noticing that the ebuild for gcc-12.3.0 got dropped with little
> explanation.  It is the upstream stable release.  I am eyeballing
> #906310 as what may have triggered the drop, but I find it a bit of a
> stretch that an upstream stable release got dropped over a single,
> optional package that has a history of quirky behavior (FWIW, I never
> had luck with ccache, especially on MIPS).

Please see https://bugs.gentoo.org/908258. There were miscompilations
even fixed after 12.3.0 was tagged.

(Also, ccache really isn't a "package with quirky behaviour" in terms of
whether or not it causes gcc to ICE. It has nothing to do with what
ccache itself does at runtime.)

>>
> Under qemu, it takes about 4 hours to build the single-ABI variant of
> gcc and 7 hours for the multilib variant.  So I avoid rebuilding the
> compiler as much as possible, as with six chroots, that's virtually an
> entire day across all six just for gcc, minus distractions (seriously,
> the build times on gcc are getting waaaaaaay out of hand, regardless
> of arch).

It should get a bit better as of recent 13 as we backported a change
to help parallel builds at least (and reduce resource consumption).

(Also, added toolchain@ to CC...)

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 377 bytes --]

Re: [gentoo-dev] What happened to gcc-12.3.0?

[Joshua Kinard]

On 6/15/2023 07:37, Sam James wrote:
> 
> Joshua Kinard <kumba@gentoo.org> writes:
> 
>> Noticing that the ebuild for gcc-12.3.0 got dropped with little
>> explanation.  It is the upstream stable release.  I am eyeballing
>> #906310 as what may have triggered the drop, but I find it a bit of a
>> stretch that an upstream stable release got dropped over a single,
>> optional package that has a history of quirky behavior (FWIW, I never
>> had luck with ccache, especially on MIPS).
> 
> Please see https://bugs.gentoo.org/908258. There were miscompilations
> even fixed after 12.3.0 was tagged.
> 
> (Also, ccache really isn't a "package with quirky behaviour" in terms of
> whether or not it causes gcc to ICE. It has nothing to do with what
> ccache itself does at runtime.)

True, I've just never had solid luck with it in the cases where I've tried using it.  Same goes for distcc. 
Something always broke, and it took more time to dig into the break and find a fix than to just build things 
the regular way.  Shouldn't be lumping a compiler ICE into that, but I wrote my inquiry late at night when I 
probably should've slept on it some more :)

>>>
>> Under qemu, it takes about 4 hours to build the single-ABI variant of
>> gcc and 7 hours for the multilib variant.  So I avoid rebuilding the
>> compiler as much as possible, as with six chroots, that's virtually an
>> entire day across all six just for gcc, minus distractions (seriously,
>> the build times on gcc are getting waaaaaaay out of hand, regardless
>> of arch).
> 
> It should get a bit better as of recent 13 as we backported a change
> to help parallel builds at least (and reduce resource consumption).

This is good to know, thanks.  I'll look at what qlop reports for build-times on the host itself as well as 
inside a qemu-mips chroot and see if there are noticeable differences.  I've been wondering if gcc upstream 
would ever look into dealing with C++'s slower compilation time in some form, so if they are starting to 
address things (beginning w/ better parallelization), that's a hopeful sign.

-- 
Joshua Kinard
Gentoo/MIPS
kumba@gentoo.org
rsa6144/5C63F4E3F5C6C943 2015-04-27
177C 1972 1FB8 F254 BAD0 3E72 5C63 F4E3 F5C6 C943

"The past tempts us, the present confuses us, the future frightens us.  And our lives slip away, moment by 
moment, lost in that vast, terrible in-between."

         --Emperor Turhan, Centauri Republic

Re: [gentoo-dev] What happened to gcc-12.3.0?

[Andreas K. Huettel]

[-- Attachment #1: Type: text/plain, Size: 745 bytes --]

Am Donnerstag, 15. Juni 2023, 06:02:14 CEST schrieb Joshua Kinard:
> 
> Noticing that the ebuild for gcc-12.3.0 got dropped with little explanation.  It is the upstream stable 
> release.  I am eyeballing #906310 as what may have triggered the drop, but I find it a bit of a stretch ...

This is for exactly the same reason as why we don't have glibc-2.36(-r0) or 2.37(-r0) in the tree anymore.
There's a stable upstream branch which accumulates bug and security fixes. The only real difference is 
naming, but I could switch to glibc-2.36_p20230615 too...


-- 
Andreas K. Hüttel
dilfridge@gentoo.org
Gentoo Linux developer 
(council, comrel, toolchain, base-system, perl, libreoffice)
https://wiki.gentoo.org/wiki/User:Dilfridge

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 833 bytes --]