* Jakub Jelinek:

> On Wed, May 27, 2026 at 10:32:08AM +0200, Dan Horák wrote:
>> On Wed, 27 May 2026 10:21:13 +0200
>> Pavol Sloboda <[email protected]> wrote:
>> 
>> > Hello,
>> > 
>> > I believe I have encountered a similar/the same issue but for me it
>> > triggers on all the architectures:
>> 
>> the icu rebuild is run with "fail_fast = True", so not all build tasks
>> are allowed to really finish, they are cancelled after a first fail
>> instead.
>
> I've managed to reproduce the boost ICE on i686, but it is really weird.
> The ICE is within mpfr:
> #0  0x00000000 in ?? ()
> #1  0xf7ee9537 in mpfr_cache (dest=0xffffaeec, cache=0xf79effd4, 
> rnd=MPFR_RNDU) at /usr/src/debug/mpfr-4.2.2-3.fc44.i386/src/cache.c:102
> #2  0xf7ec20b3 in mpfr_exp2 (y=0xffffaff0, x=0xffffaf78, rnd_mode=MPFR_RNDN) 
> at /usr/src/debug/mpfr-4.2.2-3.fc44.i386/src/exp2.c:141
> #3  0xf7ed343a in mpfr_pow (z=0xffffaff0, x=0xffffaff0, y=0xffffafe0, 
> rnd_mode=MPFR_RNDN) at /usr/src/debug/mpfr-4.2.2-3.fc44.i386/src/pow.c:665
> #4  0x09797f41 in do_mpfr_arg2(real_value*, int (*)(__mpfr_struct*, 
> __mpfr_struct const*, __mpfr_struct const*, mpfr_rnd_t), real_value const*, 
> real_value const*, real_format const*) [clone .lto_priv.0] ()
> #5  0x09797a66 in fold_const_call_sss(real_value*, combined_fn, real_value 
> const*, real_value const*, real_format const*) [clone .lto_priv.0] ()
> #6  0x0909e6a1 in fold_const_call_1(combined_fn, tree_node*, tree_node*, 
> tree_node*) [clone .lto_priv.0] [clone .cold] ()
> #7  0x0a089231 in fold_builtin_n(unsigned long long, tree_node*, tree_node*, 
> tree_node**, int, bool) [clone .isra.0] ()
> but reproduces only in f45 and not in f44, gcc is the same source
> (gcc-16.1.1-2.fc{45,44}), mpfr is even exactly the same (mpfr-4.2.2-3.fc44),
> gmp is also exactly the same.
> Reproducer can be even e.g. just
> echo 'double a = __builtin_log (60e6);' > test.c
> gcc test.c
>
> Wonder if something hasn't changed on the glibc side...

In rawhide, glibc copies the TLS initialization image before
relocation, so pointer initializations are lost.  Adhemerval has
already posted a fix:

[PATCH] elf: Re-initialise static TLS after .tdata relocation (BZ 34164)
<https://inbox.sourceware.org/libc-alpha/[email protected]>

The v2 version merely adds a bug reference.
-- 
_______________________________________________
devel mailing list -- [email protected]
To unsubscribe send an email to [email protected]
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/[email protected]
Do not reply to spam, report it: 
https://forge.fedoraproject.org/infra/tickets/issues/new

Reply via email to