On 2025-01-18 11:45, Lasse Collin wrote:
On 2025-01-18 Paul Eggert wrote:
Does the unaligned read trick work even with CheriBSD's memory-safe
model? That is an edge case that might need an ifdef or something.
I'm not familiar with CheriBSD but the trick never crosses a cache line
boundary (or
The renameatu and rename tests fail on GNU/Hurd since renameat2 on that
platform does not properly handle trailing slashes. I've reported that
bug with a test case [1].
Applied the attached patch which adds a configure check and documents
the bug.
[1] https://sourceware.org/bugzilla/show_bug.cgi?
* doc/gnulib.texi (alignof):
Improve doc to match implementationa better.
* m4/stdalign.m4 (alignas): Do not define to _Alignas
if __xlC__ claims to conform to C11, because _Alignas fails with
weird diagnostics "Unexpected text %1$s encountered."
* tests/test-alignasof.c (TEST_ALIGNMENT, alignas):
Hi Collin,
> The following check caused the linkat tests to fail on GNU/Hurd:
>
>ASSERT (linkat (dfd, BASE "link1", dfd, BASE "sub1/", 0) == -1);
>ASSERT (errno == EEXIST || errno == ENOTDIR);
Yes, I've seen that too.
> Where "link1" was an existing directory and "sub1" was an existing
* lib/flexmember.h (FLEXALIGNOF): Use the conservative definition
if _Alignof is a macro, to avoid a C99 conformance issue exposed
by IBM XL C 16.1 which otherwise complains "An aggregate
containing a flexible array member cannot be used as a member of a
structure or as an array element."
---
Chan
Hi Bruno,
Bruno Haible writes:
> Err. I disagree with that. When we have a test failure, the first step is
> to look at POSIX ([1] in this case). It says "These functions shall fail if:
> ...[ENOTDIR]" which catches this case. So, it's a bug in Hurd.
>
> Then, there are three possibilities:
>
>
Collin Funk wrote:
> I've filed the bug report in glibc under the Hurd component so hopefully
> it will get fixed [1].
Thanks. That's how we make progress.
Bruno
The following check caused the linkat tests to fail on GNU/Hurd:
ASSERT (linkat (dfd, BASE "link1", dfd, BASE "sub1/", 0) == -1);
ASSERT (errno == EEXIST || errno == ENOTDIR);
Where "link1" was an existing directory and "sub1" was an existing
symlink to a directory. Once we allow errno == E
2024-12-31 I wrote:
> I reproduce it only on Adelie Linux, not on Alpine Linux or Ubuntu, for
> example,
> because in config.h:
>
> * On Adelie Linux
> /* #undef HAVE_MINIMALLY_WORKING_GETCWD */
> /* #undef HAVE_PARTLY_WORKING_GETCWD */
>
> * On Ubuntu
> #define HAVE_MINIMALLY_WO
Lasse Collin wrote:
> About sanitizers: They can be annoying with SIMD code. If a function is
> passed an unaligned buffer, it would be fine to round the address down
> to an aligned value, do an aligned read, and ignore the out-of-bounds
> bytes. One can do it in assembly because sanitizers don't
Thanks for discussion, Lasse. I suppose the current situation is
difficult to change due to mainly licensing...
Lasse Collin writes:
> I tested Gnulib's lib/crc-x86_64-pclmul.c (v1.0-1387-g5dd298feb0) a
> bit. I changed the target attribute to plain "pclmul" to make it work
> on a processor tha
IIRC we test values from 0-16 bytes in the unit test and the pclmul
implementation doesn't get hit for smaller values?
On Sat, Jan 18, 2025, 13:16 Simon Josefsson wrote:
> Thanks for discussion, Lasse. I suppose the current situation is
> difficult to change due to mainly licensing...
>
> Lasse
On 2025-01-17 Simon Josefsson wrote:
> Lasse Collin writes:
>
> > [1]
> > https://github.com/tukaani-project/xz/blob/master/src/liblzma/check/crc_x86_clmul.h
> >
>
> Oh. Do you see any chance to merge your code into gnulib? If there
> is any feature in your implementation that we don't hav
On 2025-01-18 07:48, Bruno Haible via Gnulib discussion list wrote:
Regarding your trick to do an aligned read on (addr & -alignment) instead
of an unaligned read on (addr): I find it good that ASAN catches this,
because this trick amounts to exploiting a coincidental property of current
hardware
Paul Eggert wrote:
> Does the unaligned read trick work even with CheriBSD's memory-safe
> model? That is an edge case that might need an ifdef or something.
Good question...
> The aligned read trick reminds me a bit of the "adding 0 to a null
> pointer gives you a null pointer" trick. That als
On 2025-01-18 Bruno Haible wrote:
> Valgrind was a tool without replacement, for many years, when
> sanitizers did not exist. Nowadays, however, I generally prefer
> testing with sanitizers than with valgrind because there are some
> bugs that ASAN finds and valgrind doesn't [1]. For example, when
On 2025-01-18 Paul Eggert wrote:
> Does the unaligned read trick work even with CheriBSD's memory-safe
> model? That is an edge case that might need an ifdef or something.
I'm not familiar with CheriBSD but the trick never crosses a cache line
boundary (or page boundary). So the memory-safe model
17 matches
Mail list logo
