On 10/14/11 17:25, Eric McCorkle wrote: > I've seen the same behavior with a clang build.
Actually, no. I looked into it more, and what I found is suspicious. I had the tests failing in Magick++/attributes.sh. What's suspicious is it would claim to have a random, and often nonsensible number of failures (-21563, or something). I opened up attributes.cpp, expecting to find an uninitialized variable, however, there is an "int failures = 0" at the top of the main function. After a while, I moved the declaration outside of the main function, and the test worked. This tells me it's getting corrupted somehow, perhaps by a buffer overflow of some kind. Has anyone else seen this?
signature.asc
Description: OpenPGP digital signature