Justin C. Walker
justin at mac.com
Tue Dec 13 20:09:12 CET 2011
I am a Sage developer <http:www.sagemath.org>, and we package and build libgpg_error as part of Sage.
We recently discovered a problem with parallel builds on fast systems with a lot of cores (we have seen this on 12- and 8-core systems, both with two threads per core).
The problem appears to be in the top-level Makefile of libgpg_error: a number of "mkdir" commands (not embedded in macros, AFAICT) are invoked without a leading "-" to avoid stopping the build if the directory exists.
It can happen, if the system is fast enough and enough "make"s are going on simultaneously, that the check in the Makefile will indicate that directory does not exist, in multiple threads, and each of those threads will attempt to make the directory (with only one succeeding).
Is this a known problem that's been fixed? We're currently using version 1.6. I'm not signed up for this project, so I don't know the bug-tracking system or how to find out such things.
Justin C. Walker, Curmudgeon at Large
Institute for the Enhancement of the Director's income
What if the hokey pokey
really *is* what it’s all about?
More information about the Gnupg-devel