Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Signed-off-by: Paper <paper@paper.us.eu.org>
|
|
Signed-off-by: Alex Gaynor <alex.gaynor@gmail.com>
|
|
Needed for Adaptec ARC configuration utility.
Signed-off-by: Ariadne Conill <ariadne@dereferenced.org>
|
|
Unfortunately, we do have a potential race if the binary is renamed
between the call to readlink and the real fopen.
Fixes: #349
|
|
|
|
Used by GraalVM Java.
|
|
We'll just say we're always multi-threaded (same as glibc does in duplicate copies of itself loaded into the same process via dlmopen)
|
|
libstdc++ calls these directly
|
|
Previously, initstate_r would attempt to pack the n, i, and j
values into the state array pointed to by the x member value of
buf, which would cause access to an invalid region of memory when
buf was uninitialized. However, packing the values into the state
array was required when buf was initialized, as otherwise the
values of n, i, and j would not be saved. Critically, initstate_r
had no means by which to distinguish between an uninitialized and
initialized buf struct.
This change makes the random_r and srandom_r do the packing and
unpacking of the n, i, and j values instead. This does lead to a
somewhat more inefficient algorithm, but preserves compatibility
in case setstate_r is passed a state array that was initialized
with the normal initstate function.
Signed-off-by: George Matsumura <gm960420@ohio.edu>
|
|
|
|
|
|
This fixes a typo in commit 810f46ab1cc801e54fc3836c22b5f1364cd632e6
that created an infinite recursive call.
|
|
|
|
|
|
|
|
Prevents segfault when calling initstate_r.
|
|
|
|
Add reentrant random functions from LSB
See merge request adelie/gcompat!11
|
|
Add __fdelt_warn as a strong alias of __fdelt_chk, similar to how
it is implemented in glibc.
Signed-off-by: George Matsumura <gm960420@ohio.edu>
|
|
This adds random_r() and associated functions, mostly adapted from
musl's implementation of random().
Signed-off-by: George Matsumura <gm960420@ohio.edu>
|
|
|
|
Now newlocale calls __newlocale with the returned glibc_locale structure
directly, making it work slightly better with non-"C" locales.
|
|
* Count size of list correctly
* Don't make __names const; it isn't.
|
|
|
|
Implemented using the CPU_{ALLOC,FREE} macros.
These symbols are required by nvidia 450.57 binaries.
|
|
Ensure aarch64 bug doesn't appear here either.
|
|
|
|
Calls the more standard sched_yield().
|
|
|
|
Correct __*_finite math functions
See merge request adelie/gcompat!8
|
|
This corrects a mistake in the finite math functions in which it
was asserted that the return values had to be infinite, instead of
the intended behavior of asserting that they were not infinite.
Signed-off-by: George Matsumura <gmmatsumura01@bvsd.org>
|
|
This is similar to strtoll_l() in a way it ignores
the locale_t argument and uses the global locale
settings.
Signed-off-by: Michal Kazior <kazikcz@gmail.com>
|
|
This is similar to strtoll_l() in a way it ignores
the locale_t argument and uses the global locale
settings.
Signed-off-by: Michal Kazior <kazikcz@gmail.com>
|
|
This is similar to strtoll_l() in a way it ignores
the locale_t argument and uses the global locale
settings.
Signed-off-by: Michal Kazior <kazikcz@gmail.com>
|
|
Signed-off-by: Michal Kazior <kazikcz@gmail.com>
|
|
Signed-off-by: Michal Kazior <kazikcz@gmail.com>
|
|
|
|
|
|
Newer releases of Chromium directly execv /proc/self/exe, missing our
readlink(3) interposing. We go ahead and interpose all three execv*
functions just to ensure wider compatibility.
|
|
The assertion should be that the return value was 0.
Additionally, show the result code if the return value is not 0.
|
|
This is necessary because the musl and glibc function signatures
differ significantly.
|
|
|
|
|