building rust infinite loops in __memset_sse2 on Indiana University network machines (RHEL 6.4 x64)

[toddaaro]

This is a mysterious issue. Something about the configuration of the Indiana University network results in rustc infinite looping during make. This has been reproduced on three different machines and two user accounts. Attaching gdb to the running rustc process spots a number of threads waiting and one in __memset_sse2. So far every (10+) gdb attach has had the same memset call on top of the backtrace.

Here are a few pastes that might be useful. The first is a configure + check-stage0-std output. The second is gdb attached and used to get a pair backtraces.

pastebin.mozilla.org/2884116
pastebin.mozilla.org/2884115

The machine configurations are 64 bit Redhat Linux Enterprise Edition 6.4 (Santiago). Glibc is version 2.12, and gcc is version 4.4.6-3 Redhat.

[toddaaro]

Still haven't followed up on this. I might not get to it until late next week.

[alexcrichton]

Due to lack of reproducability and the pastebins being gone now, I'm going to close this. It'd be tough for us to even verify whether a fix worked or not sadly.