I'm currently trying to build a patch for my -no-static proposal (see my previous e-mail).
So the first thing I did was to perform a cvs checkout of libtool and tried a 'make' and (for the first time) a 'make check'. The 'make check' resulted in a total process starvation on my system for the first test it performed. After some digging I found out the reason:
- when you issue a 'make check', first a ./configure is performed in the cdemo directory.
- one of the tests that is performed by configure is to verify whether for fort77 (which is available on my system) the PIC flag -fno-common works.
- it seems that when fort77 is called to compile conftest.f the environment variable 'CC=fort77' is set.
- Since fort77 is a script which uses the CC environment variable to call the C compiler the script now starts calling itself recursively and therefore totally starves my system of process resources.
Strangely enough the CC=fort77 environment setting does not appear to the fort77 script when you call './configure' from the command prompt. It only appears when ./configure is called via a 'make test'. I was not able to find out what precisely triggers the passing of the CC environment variable when '(eval "$lt_compile" 2>conftest.err)' is called, but it sure seems like a bug of some sort to me that this can occur.
FYI, I'm using the following:
platform: Mac OS X 10.2.6
autoconf: 2.57
automake: 1.7.6
libtool (installed version): 1.5 (1.1220 2003/04/05 19:32:58)
libtool (cvs version, on which I performed the 'make check'): 1.5a (1.1243 2003/07/15 00:05:21)
fort77: latest version as retrieved through fink
Regards, Sander Niemeijer
_______________________________________________ Libtool mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/libtool
