jemalloc 4.0.2 released
jasone at canonware.com
Wed Sep 23 12:23:30 PDT 2015
The test failure is harmless, but I'll change the test to allocate an additional maximal object if the first one succeeds; two objects each more than half the virtual address space cannot possibly coexist, so the second allocation must OOM.
> On Sep 22, 2015, at 6:02 AM, Ingvar Hagelund <ingvar at redpill-linpro.com> wrote:
> build check of jemalloc-4.0.2 fails on i686, on RHEL, (el5 and el6
> fails, el7 does not have i386) and fedora up to f22.
> === test/integration/mallocx ===
> test_oom:test/integration/mallocx.c:57: Failed assertion:
> (mallocx(hugemax, 0)) == (NULL) --> 0x17000000 != 0x0: Expected OOM for
> mallocx(size=0xe0000000, 0)
> test_oom: fail
> Example build check fail (they all fail on the same check):
> Note that on f23/i686, the build check completes without the assertion
> Den 21. sep. 2015 21:03, skrev Jason Evans:
>> jemalloc 4.0.2 is now available. This bugfix release addresses a few bugs specific to heap profiling.
>> Bug fixes:
>> - Fix ixallocx_prof_sample() to never modify nor create sampled small allocations. xallocx() is in general incapable of moving small allocations, so this fix removes buggy code without loss of generality.
>> - Fix irallocx_prof_sample() to always allocate large regions, even when alignment is non-zero.
>> - Fix prof_alloc_rollback() to read tdata from thread-specific data rather than dereferencing a potentially invalid tctx.
>> For the complete ChangeLog, see:
>> Direct download:
>> Starting point for general information:
>> Browsable revision history:
>> jemalloc-announce mailing list
>> jemalloc-announce at canonware.com
> jemalloc-discuss mailing list
> jemalloc-discuss at canonware.com
More information about the jemalloc-discuss