Segfault (probably due to stack overflow) from 6g. #1617
Description
by consalus:
With weekly.2011-03-15 (and tip) on darwin and linux 6g, when I try to compile the "gotcl" project, the compiler segfaults, apparently due to stack overflow. When run in gdb, it looks like the issue occurs in sinit.c. The stack frames show bouncing back and forth between init2 and init2list, though the eventual segfault appears to happen in init1. My guess would be a cycle in the Node tree. This issue doesn't appear to happen with the "release" branch or the "weekly.2011-03-07.1" branch, so I assume it was introduced sometime between the 7th and the 15th. I haven't gotten a chance to narrow the change range further. Steps to duplicate: hg clone https://gotcl.googlecode.com/hg/ gotcl cd gotcl make For me on two machines, that results in a segfault. Interestingly enough, if one changes "chans.go" by commenting out line 136, there is no segfault and everything compiles normally and tests pass. This phenomenon doesn't appear to be dependent on the position of "forchan". When "forchan" is added to the map by other means (in init), everything compiles and it works fine.