Expected Behavior of Test throwto003
marlowsd at gmail.com
Mon Apr 19 08:26:36 EDT 2010
On 15/04/2010 23:21, Gabor PALI wrote:
> Based on my experiences with the nightly builds and tests, it seems that
> 2 + 4 GB (real + swap) memory is not enough sometimes :) There is a
> process (throwto003, test 173) during running `make -C
> testsuite/tests/ghc-regress BINDIST=YES` (phase 11 in the builder) which
> eats up a lot of memory and then killed by the operating system
> (FreeBSD/amd64 8.0-RELEASE) due to lack of swap space.
> Is it intentional (I assume no as it is listed as an "unexpected
> failure")? I created a brief memory profile , recorded (one snapshot
> per second) while make(1) was running.
It's supposed to run in very little memory, if it doesn't that's a bug.
Is it displaying this behaviour for all the ways? Try
$ cd testsuite/tests/ghc-regress/concurrent/should_run
$ make TEST=throwto003 WAY=normal
and substitute 'normal' for 'threaded1' and 'threaded2' to try other ways.
More information about the Cvs-ghc