On May 4, 2016, at 6:12 AM, Ingvar Hagelund <[email protected]> wrote: > test/integration/mallocx fails on older i686. It fails on el5, el6, > though it works on fedora 22. > > el5 and el6 dies at the same spot: > > === test/integration/mallocx === > test_overflow: pass > test_oom: pass > test_basic: pass > test_alignment_and_size:test/integration/mallocx.c:170: Failed > assertion: (ps[i]) != (NULL) --> 0x0 == 0x0: mallocx() error for > alignment=33554432, size=100663291 (0x5fffffb) > test/test.sh: line 22: 24762 Segmentation fault (core dumped) ${t} > /builddir/build/BUILD/jemalloc-4.1.1/ /builddir/build/BUILD/jemalloc- > 4.1.1/ > Test harness error > make: *** [check_integration] Error 1 > > Full build log for el5: https://ingvar.fedorapeople.org/jemalloc/jemall > oc-4.1.1.el5.i686.build.log
Is it possible that ~128 MiB is simply too much to allocate on those systems? Does 4.1.0 also exhibit this behavior? Thanks, Jason _______________________________________________ jemalloc-discuss mailing list [email protected] http://www.canonware.com/mailman/listinfo/jemalloc-discuss
