> Steve Hay wrote on 04 January 2005 11:01: > > >I was wondering if anybody had any suggestions, it seems > odd that with > the > > >Perl Malloc i run out of memory when without it I dont. > > > > > Perl's malloc() requires an sbrk() function, which Win32's > CRT doesn't > > provide. Therefore, Perl's win32.c source provides one, > but this had > > some limitations which would lead to the kind of out of > memory error > > that you've seen. > > > > I had a patch committed which should fix the problem that you had: > > > > http://public.activestate.com/cgi-bin/perlbrowse?patch=23128 > > > > This patch went into 5.8.6, so please retry with that and > let us know > > how you get on. > The build with 5.8.6 as you describe worked fine, but with > one caveat. You _have_ to turn DEBUG_MSTATS on otherwise it hangs on > t/Benchmark for some reason. The test number it shows as being the last one is 65 > or 79 with no obvious reason why one or the other should fail.
And when the Benchmark test is run without the harness it performs as expected. I tried patching malloc.c to move anything that modfied 'goodsbrk' out of any DEBUGGING_MSTATS ifdef blocks (on the grounds its used outside of them anyway) but that didnt appear to help. (Although im not such an experienced C programmer to say for sure i did it right.) To make things even more confusing when I run the harness on JUST that test file it all works fine. Cheers, Yves
