On 10/01/07 09:07, Patrick Ale wrote:
I don't have a ~/.make.machines file. When I look in the opensolaris.sh environmental file I see there is some calculation done, CPUs + 2 or something (which comes to 4 in my case). I think I fixed it though, outcome is I really needed more swap.. I BFU-ed to the debug archives of b73 (downloaded from the /on/downloads/ site) and tried to recompile onnv_74, same /tmp full errors, which kinda ruled out a problem with onnv_74 itself. So, compiling any onnv_gate on SXCE72 works, compiling onnv_73 and onnv_74 on b73-debug and onnv_74 debug does not work with 512MB of swap. When I repartitioned my disk and defined a 2GB slice for swap, iI was capable of compiling onnv_74 under a debug-enabled host (in this case b73) Could it be that a system with 'debug' enabled makes me hit some heavy swap or /tmp usage during compilation that I don't notice when compiling on a non-debug host? And if yes, is there a recommended value for swap or /tmp when you compile debug enabled onnv-gates? Obviously 2GB of RAM isnt enough, at least not in my case.
DEBUG kernels use a little more memory but not massive amounts, but I guess it could be enough to tip things if you were already close to the edge. Note that build environments tools such as bldenv usually set TMPDIR=/tmp which will place the burden on swap/memory - nice and fast if you have the memory, but effectively disk based if you don't. You can add more swap, as you did, or as an experiment set TMPDIR to /var/tmp or somewhere that has lots of space. Do you know which processes ran out of memory - 'cc' etc during build, or did they succeed and the lint check fail (assuming you were building with nightly which does a lint check at the end of the debug build). Lint uses a good chunk of memory and perhaps more in Studio 12 if they have added more features etc, but I would not expect a modest 4-way parallelism to exhaust memory during compile. Cheers Gavin
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ opensolaris-code mailing list opensolaris-code@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
