[Python-Dev] tests failing in 2.2.2 (original) (raw)

Neal Norwitz neal@metaslash.com
Tue, 8 Oct 2002 17:06:39 -0400

On Tue, Oct 08, 2002 at 04:25:06PM -0400, Guido van Rossum wrote:

[Tim] > > with a four-byte pointer, you attempt to allocate 0x20000000L items (due > > to roundupsize), which is small enough to try a realloc. In turn, it > > tries to allocate 0x80000000L bytes, i.e. 2GB. On a 32-bit machine, it > > is possible to allocate that much memory. > > Depending on platform, of course. Under MSVC6, it does call the platform > realloc(), and it's the latter that returns a NULL pointer (the max you can > alloc under Win32 is 0x7ffdefff). > > If we changed the test to use maxint // 2 instead, it would (just barely) > trigger the > > newsize <= ((~(sizet)0) / sizeof(type))_ _> > test in NRESIZE instead, and keep the platform realloc() out of it. > > Good enough?

[Guido] I'd say so, except I fear that it might cause different problems (not that I know of any). I tried it on Linux and on Win98 and it works fine. I presume you've tested it on Win2K. But wasn't it specifically designed to catch a condition that was only true when the size was (approximately) maxint/4?

The original problem was actually 1e9. The test is somewhat dubious, so anything to reduce the problems would be good. Perhaps the test should just be removed?

Neal