At 12:05 PM 6/25/2007, Christian Convey wrote:
So what puzzles me is that I was seeming to get a build error that was
shell-specific. Yet I see plenty of information from man pages, folks
like you, etc. that say I should never be seeing shell-specific build
errors.
You're asking people to remotely diagnose an anecdotal problem which
you haven't even tried to reproduce, and I don't think that's fair.
If you want to pursue this, put back the broken syntax and try to
reproduce the problem with a csh user. Once you have a test case we
(or I at least) will be happy to engage. If you're on Solaris, start
diagnosis by prepending "truss -texec -f -a --" to the make command.
If you're on Linux, strace will do a similar job. If you find a csh
being exec-ed by make, we'll have some meat to chew on.
-David Boyce
PS This most likely is related to the fact that bash has partial
support for csh compatibility. Specifically the ">&" syntax seems to
work in bash where it wouldn't in a POSIX or Bourne shell. As for the
difference between your experience and other users, maybe bash
repented of this hack and removed it in a later version. Also, it
behaves differently dpeending whether it was run as /bin/bash or as
/bin/sh (the intent is that it be more POSIX- conformant as /bin/sh)
so maybe there's a difference there. Or you may have the
POSIXLY_CORRECT environment variable defined while others don't.
There are many possibilities but "GNU make defaults to the user's
shell" is not one of them.
_______________________________________________
Help-make mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-make