Paul Eggert wrote:
> It's probably simpler to have the openat replacement return the lowest
> available fd as per POSIX. I installed the attached untested patch to
> try to do that, and we can see whether CI likes it.
Thanks! This fixed it: the CI build of libunistring now succeeded.
Bruno
On 2024-11-11 02:06, Bruno Haible wrote:
The test, intest-openat.c:94, is reasonable IMO: It verifies that the
openat() replacement for native Windows behaves the same as a POSIX-compliant
openat()
What do you think? Should some module dependency be added to the 'openat'
module? Or to the 'sa
On 11/11/2024 16:47, Paul Eggert wrote:
On 2024-11-10 05:48, Pádraig Brady wrote:
BTW I've pushed a tweak to gnulib to avoid a -Werror=unused-variable
issue with --disable-acl
Thanks, I installed the attached further patch, since the res5t of the
file uses MAYBE_UNUSED.
Thanks for all the fi
Paul Eggert wrote:
> diff --git a/modules/opendirat b/modules/opendirat
> index 17542ff4fa..61500512ee 100644
> --- a/modules/opendirat
> +++ b/modules/opendirat
> @@ -10,7 +10,7 @@ c99
> dirent
> fcntl-h
> fdopendir
> -openat-safer
> +openat
>
> configure.ac:
This change triggered a build f
On 2024-11-11 07:58, Paul Eggert wrote:
For the Cygwin issue, I am easily confused
by "else" immediately followed by "#endif" or "#else" or "#elif", and so
installed the attached further patch which is merely a refactoring.
Come to think of it, we also should use acl_get_link_np later in that
* lib/malloca.c (freea) [__CHERI_PURE_CAPABILITY__]:
Add a cast as the code otherwise doesn’t conform to C.
---
ChangeLog | 4
lib/malloca.c | 2 +-
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/ChangeLog b/ChangeLog
index abeae1e6b6..05594a9581 100644
--- a/ChangeLog
+++
Hi,
Markus Mützel wrote:
> Octave is using gnulib's "fseek" and "ftell" functions. A user reported that
> they have issues when using these functions with large files on Windows:
> https://savannah.gnu.org/bugs/index.php?66399
This is explained in the Gnulib documentation:
https://www.gnu.org/so
On 2024-11-10 05:48, Pádraig Brady wrote:
BTW I've pushed a tweak to gnulib to avoid a -Werror=unused-variable
issue with --disable-acl
Thanks, I installed the attached further patch, since the res5t of the
file uses MAYBE_UNUSED.From 6a018d0492239d01c2cc8fd56a1acec4d6fcd44d Mon Sep 17 00:00:0
On 2024-11-10 22:25, Bruno Haible wrote:
../../gllib/file-has-acl.c:427:21: error: 'acl_get_link_np' undeclared (first
use in this function)
427 | : acl_get_link_np)
| ^~~
Thanks for fixing those bugs. For the Cygwin issue, I am easil
Be able to get the security context even if configured with
--disable-acl, as security contexts are not ACLs, and the
main reason for --disable-acl was for efficiency with GNU ls -l,
a concern that does not apply to security contexts (which are
needed only with ls -Z). Problem reported by Pádraig
Florian Weimer wrote:
> * lib/nproc.c (num_processors_via_affinity_mask): Retry
> with larger affinity masks if CPU_ALLOC_SIZE is available.
Thanks, applied.
To improve the fallback case, I added this patch:
2024-11-11 Bruno Haible
nproc: Use affinity mask even in out-of-memory situ
Octave is using gnulib's "fseek" and "ftell" functions. A user reported that
they have issues when using these functions with large files on Windows:
https://savannah.gnu.org/bugs/index.php?66399
If I understand the code paths for Windows correctly, the issue might be caused
by the fact that the
* Bruno Haible:
> Hi Florian,
>
> Thanks for the patch.
>
> Florian Weimer wrote:
>> +#elif HAVE_SCHED_GETAFFINITY_LIKE_GLIBC \
>> + && defined CPU_ALLOC_SIZE /* glibc >= 2.6 */
>> + {
>> +unsigned int alloc_count = 1024;
>> +while (1)
>> + {
>> +cpu_set_t *set = CPU_ALLOC (
* lib/nproc.c (num_processors_via_affinity_mask): Retry
with larger affinity masks if CPU_ALLOC_SIZE is available.
---
v2: Handle CPU_ALLOC_SIZE failure.
lib/nproc.c | 28
1 file changed, 28 insertions(+)
diff --git a/lib/nproc.c b/lib/nproc.c
index 92a07e8289..48bc3
Hi Florian,
Thanks for the patch.
Florian Weimer wrote:
> +#elif HAVE_SCHED_GETAFFINITY_LIKE_GLIBC \
> + && defined CPU_ALLOC_SIZE /* glibc >= 2.6 */
> + {
> +unsigned int alloc_count = 1024;
> +while (1)
> + {
> +cpu_set_t *set = CPU_ALLOC (alloc_count);
According to the
* lib/nproc.c (num_processors_via_affinity_mask): Retry
with larger affinity masks if CPU_ALLOC_SIZE is available.
---
lib/nproc.c | 26 ++
1 file changed, 26 insertions(+)
diff --git a/lib/nproc.c b/lib/nproc.c
index 92a07e8289..14da5f6a9b 100644
--- a/lib/nproc.c
+++ b/
This week's CI of GNU libunistring has a new test failure, not present
last week, on native Windows (mingw, MSVC):
FAIL: test-openat
=
../../tests/test-openat.c:94: assertion 'openat (dfd, ".", O_RDONLY) ==
STDIN_FILENO' failed
FAIL test-openat.exe (exit status: 3)
The o
The functions rpl_acl_get_fd, rpl_acl_set_fd are part of the
'acl-internal' module, since they are declared in acl-internal.h.
Their definition should be in the same module. Currently they
are in qset-acl.c, which makes no sense, because it is possible
to use the 'acl-internal' module without the '
The getpagesize() function should better follow the stated rule for
use of 'extern inline'. Otherwise we could get into trouble like for
realloc-posix.
2024-11-11 Bruno Haible
getpagesize: Fix use of 'extern inline' (regression 2012-11-20).
* lib/unistd.in.h (_GL_GETPAGESIZE_I
On 2024-11-05 I wrote:
> This patch fixes it. It is possible that the same bug still occurs when there
> are 2 or more gnulib-tool invocations in the scope of the same configure file;
> still need to think about this case.
Indeed, the problem was not completely fixed: The CI reported a GNU libunis
20 matches
Mail list logo
