Re: [jamboost] spawn error?

From: Rene Rivera (grafik.list_at_[hidden])
Date: 2004-03-05 14:07:11

Beman Dawes wrote:

> bjam quit early on today's Win32 regression tests. Pavol Droba had added 
> three additional tests. On the first one, bjam did OK for borland and 
> cwpro9 toolsets, and then aborted on the gcc toolset. See below.
> 
> Any ideas?
> 
> Even if the problem is caused by Pavol messing up with his Jamfile, it 
> would be better if bjam issued an error message rather than just aborting.
> 
> --Beman
> 
> MkDir1 
> d:\boost-regr\bin\boost\libs\algorithm\string\test\container.test\gcc
> MkDir1 
> d:\boost-regr\bin\boost\libs\algorithm\string\test\container.test\gcc\debug
> gcc-C++-action 
> d:\boost-regr\bin\boost\libs\algorithm\string\test\container.test\gcc\debug\container_test.obj
> gcc-Link-action 
> d:\boost-regr\bin\boost\libs\algorithm\string\test\container.test\gcc\debug\container.exe
> spawn: No such file or directory

I'm not sure how the additional tests would cause problems. Or how it would 
work for one toolset and not another. But I did make changes to bjam and 
Boost.Build last night to fix a spawn related problem.

gcc-tools.jam changed to add some missing quotes around the commands. 
Specifically archive and link did not have them and would fail if the command 
had spaces in it (for example when a compiler is installed in \Program 
Files\..). This uncovered a problem in the NT spawn which doesn't accept 
quotes around the command name when we invoke it directly, rather than using 
CMD.EXE, which is what the gcc link step does.

So basic question is.. did you rebuild bjam from CVS for the tests? If not, 
that's likely the cause of the problem as the gcc-tools now use quotes around 
the commands, but the older (before last night) bjam can't handle.

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com - 102708583/icq