- https://groups.google.com/forum/#!topic/mozilla.dev.tech.nspr/6yPZqtZhHzM
- remove rwlockrank.c patch, merged in #1072454
- remove chunk of patch-nspr_pr_src_pthreads_ptthread_c, merged upstream
in #756047. Leave chunk about pthread_set_name_np() returning void, to
be upstreamed
- use PATH_MAX instead of MAXPATHLEN in patch-nspr_pr_src_linking_prlink_c
NSPR headers define OPENBSD (which is what we need to know what OS we're
on), sys/param.h defines OpenBSD. And MAXPATHLEN, which is used in that
patch. Oh well, fix it this way for now, pending a better fix.
Might or might not fix security issues, noone knows as there are no
release announcements, nor changelogs. Check
http://hg.mozilla.org/projects/nspr/shortlog for a list of changes...
Fixes the following bugs:
Bug 927687: Avoid unsigned integer wrapping in PL_ArenaAllocate (possible CVE?, not confirmed yet)
Bug 770534: eliminate possible pointer overflow in PL_ArenaAllocate().
Bug 888546: PR_ImportUDPSocket doesn't work.
No release notes yet.
Note: This is what will be shipped within Fx 25.0.1, per #935568 & #935959.
http://mozilla.6506.n7.nabble.com/ANNOUNCE-NSPR-4-10-Release-td280660.html
After years of sticking to CVS, upstream finally switched to hg,
changing its directory layout while here. All our patches are actually
unchanged, but got renamed/moved. CVS doesn't handle moves :)
Tested in an amd64 bulk build.
NSPR 4.9.6 has the following bug fixes:
* Bug 782214: Auto detect Android API version for x86.
* Bug 807883: Add the PL_SizeOfArenaPoolExcludingPool function.
* Bug 812085: Initialize Windows CRITICAL_SECTIONs without debug info
and with nonzero spin count.
* Bug 827271: Add support for Aarch64 (64-Bit ARM).
* Bug 841651: PR_SetThreadPriority doesn't do anything when using Bionic.
Bump minor for PL_SizeOfArenaPoolExcludingPool addition.
After two years of discussion in bug #634793 and countless hours lost on
filing followup bugs at each type mismatch, nspr upstream finally got
convinced that on OpenBSD uint64_t was unsigned long long, and there was
no reason at all to define PRUint64 as unsigned long, happily raping the
principle of least surprise. So now that upstream behaves correctly,
remove the two pr{long,types}.h patches.
And there was much rejoicing.
Tested on sparc64/ppc/hppa/amd64, went in a bunch of bulk builds.
- remove subpackages, it makes no sense to have -docs fetching its distfile
from a separate tarball. If really needed one can make a standalone port.
- remove handmade nspr.pc.in, upstream provides one
- remove handrolled install target and use the one upstream provides.
- merge PFRAG.shared back into PLIST
Tested on amd64/i386/powerpc and in a bulk build.
Needed by firefox 18.
ok sthen@ ajacoutot@ jasper@
- Garbage collect patch adding -pthread hack while here.
- Improve pthread_set_name_np() call (returns void)
- Needed anyway by upcoming Mozilla 16 releases
been in a bunch of amd64 bulks, ok ajacoutot@
- using /dev/arandom instead of /dev/urandom is not needed anymore
- #undef _POSIX_THREAD_PRIORITY_SCHEDULING is not needed since the
switch to rthreads.
from brad.
Tested on alpha amd64 hppa i386 powerpc sparc64, no change in regression
output (well, dtoa still fails, but thats not surprising given
corresponding code is heavily patched)
.so.x.y passed to PR_LoadLibrary() is never the same as the one on
linux, since the portstree controls the .x.y, so if the first dlopen()
fails, let's strip what's after .so and retry, since ld.so is clever
enough to open the correct lib if asked for libfoo.so. Allows me to
remove smth like 40 patches in all mozilla ports. Take maintainership
while here.
discussed with deraadt@
- PL_ArenaFinish should zero the static variable PRCallOnceType
once.
- Missing _MD_UNLOCK() in IsFileLocal in ntio.c.
- PR_Interrupt() interrupts PR_TransmitFile() while PR_BlockInterrupt()
in effect.
- use /dev/urandom instead of /dev/random
ok naddy@
modes for header files in the install target, so that the resulting
package is right, and non-root/wheel users are able to build apps
depending on nspr
ok kurt@, naddy@, pval@
in -stable has internal nspr and will cause issues for users who upgrade if
SO_VERSION reaches 19.0 due to the .libs package. discussed with martynas.
okay pvalchev@
- increase soft limits for regression tests and make sure the prpollml
test works (from kili@ and kurt@)
- enable nameshm1 regression test (disable ClientServerTest function)
ok kurt@
