Paul Eggert did:
> Fix the test by putting ‘volatile’ at the right place.
Thanks!
GCC 14 on x86-64 with -O2 apparently outsmarts our test for null
pointer dereference, and this is something the C standard allows.
Fix the test by putting ‘volatile’ at the right place.
* tests/test-sigsegv-catch-stackoverflow2.c (null_pointer):
Make it a volatile pointer, not a pointer to volatile
Building a testdir with clang + UBSAN + ASAN, I see a test crash:
$ ./test-cache
AddressSanitizer:DEADLYSIGNAL
=
==3658512==ERROR: AddressSanitizer: SEGV on unknown address 0x7faa9986fff8 (pc
0x55d0bf01f1c5 bp 0x7ffe85347200 sp 0x7f
With clang 19, I see a warning:
../../gllib/unictype/scripts.h:2637:26: warning: implicit conversion from 'int'
to 'short' changes value from 32768 to -32768 [-Wconstant-conversion]
The cause is that through the Unicode 16 upgrade, the level3 array in the
generated scripts.h grew from 248 blocks
Bruno Haible wrote:
> KO Myung-Hun wrote:
* lib/wait-process.c (klibc_waitpid) [kLIBC]: New function.
(waitpid) [kLIBC]: Define it to klibc_waitpid.
>>>
>>> The code looks good, but lacks comments. In the ChangeLog or git commit
>>> entry, please state *what* has changed, not *why* or
KO Myung-Hun wrote:
> >> * lib/wait-process.c (klibc_waitpid) [kLIBC]: New function.
> >> (waitpid) [kLIBC]: Define it to klibc_waitpid.
> >
> > The code looks good, but lacks comments. In the ChangeLog or git commit
> > entry, please state *what* has changed, not *why* or *how* the code works.
>
On 9/20/24 06:21, Collin Funk wrote:
It appears that the only difference is the ./configure check.
Wasn't it more due to a license issue of pulled-in modules, GPL vs. LGPL?
Have a nice day,
Berny
