Bug#926780: unblock gcc-8/8.3.0-7 and updated cross builds
Control: tags -1 - wontfix
Control: reopen -1
On 20.06.19 13:54, Paul Gevers wrote:
> Control: tags -1 wontfix
> Control: close -1
>
> Hi Matthias,
>
> On 06-06-2019 12:01, Paul Gevers wrote:> doko, I know you are
> maintaining quite some key packages, so extra work> is probably not what
> you are looking for, but neither are we and on top> of that, we don't
> like turning down unblock request (hence the time it> took to reply, at
> least that's the reason for me). In this case, and> also for gcc-7
> (hence cc of that bug) it would be great if we could> understand from
> the beginning why you believe why we should except this.> And no, I am
> not going to find upstream repositories and bug trackers> for all the
> packages that we get unblock requests for. You'll have to> help us
> making the judgment.
> I take the lack of reply from your side to mean that you are not
> pursuing to drive this further to an unblock. To clean up the unblock
> that probably will not see any further action, I am closing it as
> wontfix. If you are still interested, you can of course reopen, but be
> aware that the time for unblocks for buster is running out quickly now [1].
well, this has a bad smell. First not replying for a month, then turning it down
after not even a week. Apologies for not immediately replying to your email.
Here are the list of changes for gcc-8. Most of them regressions found during
the development of GCC 9, and backported. These come with test cases in the GCC
testsuite, and the added tests pass. From my point of view it's safe to ship
these in buster. The changes in detail are:
PR target/89877 (ARC), target not affecting Debain
PR target/84369 (PPC), fix test on POWER9
PR tree-optimization/85762, wrong code regression in GCC 8
PR tree-optimization/87008, missed optimization, regression in GCC 8
PR tree-optimization/85459, missed optimization, regression in GCC 8
PR target/87532 (PPC), wrong code fixes, not marked as regression
PR ipa/89693, ICE on valid code, regression in GCC 9, backported to 8
PR middle-end/88587, ICE on valid code, backported to 8
PR tree-optimization/90018, x86 AVX512, wrong code, regression in GCC 7/8
PR target/90024 (ARM), ICE on valid code on ARM32, regression in 7/8
PR target/89945 (ARM), ICE on valid code, regression in 7/8/9
PR fortran/87352, compile-time-hog, memory-hog, regression in 7/8/9/10
PR fortran/89981, rejects valid code, regresion in GCC 8
PR fortran/89904, ICE on valid code, regression in GCC 9, backported to 8
PR libgfortran/79540, test failure, regression in 7/8 (PARISC only?)
PR fortran/87127, rejects valid code, not marked as regression
PR rtl-optimization/87979, ICE on valid code, not marked as a regression
PR rtl-optimization/84032. ICE on valid code, not marked as a regression
Non-upstream patches:
* Fix PR c++/90050, always link with libstdc++fs.a. LP: #1824721.
Not really needed for buster, has only effect with GCC 9 installed.
* Fix PR bootstrap/87338 on ia64 (James Clarke). Closes: #927976.
Fixes the build on ia64. Now upstreamed as well.
There are no changes for the cross packages. They are rebuilt for this gcc-8
upload. Currently we don't have matching native and cross compilers in buster.
Matthias
Reply to: