Bug#1104745: gcc-15 ICE compiling linux kernel 6.14.5 with CONFIG_RANDSTRUCT
On Wed, Jun 04, 2025 at 10:43:11PM +0200, Uwe Kleine-König wrote:
> Control: tag -1 + fixed-upstream
> Control: forwarded -1 https://lore.kernel.org/r/20250530221824.work.623-kees@kernel.org
>
> Hello,
>
> On Mon, May 05, 2025 at 05:40:57PM +0200, Ingo Saitz wrote:
> > When compiling the linux kernel (tested on 6.15-rc5 and 6.14.5 from
> > kernel.org) with CONFIG_RANDSTRUCT enabled, gcc-15 throws an ICE:
> >
> > arch/x86/kernel/cpu/proc.c:174:14: internal compiler error: in comptypes_check_enum_int, at c/c-typeck.cc:1516
> > 174 | const struct seq_operations cpuinfo_op = {
> > | ^~~~~~~~~~~~~~
>
> This is claimed to be fixed in upstream by commit
> https://git.kernel.org/linus/f39f18f3c3531aa802b58a20d39d96e82eb96c14
> that is scheduled to be included in 6.16-rc1.
I can confirm applying the patches
e136a4062174a9a8d1c1447ca040ea81accfa6a8: randstruct: gcc-plugin: Remove bogus void member
f39f18f3c3531aa802b58a20d39d96e82eb96c14: randstruct: gcc-plugin: Fix attribute addition
fixes the compile issue (on vanilla 6.12, 6.14 and 6.15 kernel trees;
the kernels seem to run fine, too, so far). The first patch was needed
for the second to apply cleanly. But I can try to backport only
f39f18f3c3531aa802b58a20d39d96e82eb96c14 and see if it still compiles.
>
> This wasn't explicitly marked for stable, but I think a backport would
> be good.
>
> Best regards
> Uwe
Ingo
--
const_cast<long double>(Λ)
Reply to: