This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Does running a parallel "make -jN check" work safely?
- From: "Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>
- To: lucier at math dot purdue dot edu
- Cc: gcc at gcc dot gnu dot org
- Date: Mon, 1 Mar 2004 10:44:05 -0500 (EST)
- Subject: Re: Does running a parallel "make -jN check" work safely?
- References: <BD72EE2B-6568-11D8-8588-003065BA681E@math.purdue.edu>
> I have not had any problems with "make -j 12 -k check" and this is
> what I use. I seem to remember a thread where some test files were
> renamed to eliminate some of these name conflicts.
>
> Brad
An update on parallel "make check", After about a week of doing it,
I'm regularly seeing one or two completely random failures in the
testsuite runs. They come up in the various C and C++ testcases, and
make the report useless. I have no idea why it's happening. :-(
The failures don't appear to be visible compiler crashes, SEGVs or
fatal errors. Instead they seem to be failure to see an expected
error message or sometimes the tests simply "FAIL" with no indication
in the logfile as to what went wrong. The random failures occur with
GCC 3.3, 3.4 and mainline branches, so I'm pretty sure it's not a real
bug in GCC, but rather some interaction with parallel "make check" and
the tools I use. To me that would point a finger at the
dejagnu/tcl/expect trio.
Brad - I'd like to figure out what if anything is different between my
setup and yours to hopefully narrow down what's wrong. I run on
sparc-sun-solaris2.7 (and sparc64) and I'm using dejagnu-1.4.3,
tcl8.4.4, expect-5.39 and make-3.79.1. Would you please let me know
what you are using? (I'm hoping it's a bug in some version of a tool
I've got.)
Thanks,
--Kaveh
--
Kaveh R. Ghazi ghazi@caip.rutgers.edu