This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [regress@cygnus.com: 3 GCC regressions, 2 new, with your patch on 2001-07-13T17:30:03Z.]


In article <200107132055.NAA05019@geoffk.org>,
Geoff Keating <geoffk@geoffk.org> writes:

> The log file says:

> [...]/g++.robertl/eb130.C:9:20: hash_set: No such file or directory
> [...]/g++.robertl/eb130.C:10: syntax error before `;' token

> which I expect means that the 'hash_set' header file is not being built.

I am sorry I didn't catch this problem, since it appeared in my tree
as well.  I confess that I only reran libstdc++'s test suite.

That header is being built correctly but ext/ has been (again) dropped
from the header directory path with the jumbo configuration patch.

On 7/11, this was the command line:

spawn /usr/users/rittle/tmp/gcc-build-latour-mainline-0711/gcc/testsuite/../g++ 
[...] -I/usr/users/rittle/outside-cvs-src/gcc-mainline/libstdc++-v3/include/ext
[...] g++.robertl/eb130.C 

After the patch, it no longer included a reference to the ext/
directory (in either the source or the staging area):

; ccvs diff -r1.4 -r1.5 testsuite_flags.in
Index: testsuite_flags.in
[...]
35,37c35
[...]
39,40c37
<               -I${SRC_DIR}/testsuite
<               -I${SRC_DIR}/include/backward -I${SRC_DIR}/include/ext"
---
>               -I${SRC_DIR}/testsuite"

Do we add back in an explicit reference to ext (updated to point to
the staged side of things)?  Or do we modify the test case?

Modifying the test case ensures that we test the bulk of the tests
against the actual headers seen once the compiler is installed.

Regards,
Loren


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]