>>> Test phase [check]: sys-devel/flex-2.5.4a-r5
./flex -t -p ./scan.l \
| sed s,\"./scan.l\",\"scan.l\", \
| diff scan.c -
14a15
>
1482c1483
< register char *yy_cp, *yy_bp;
---
> register char *yy_cp = NULL, *yy_bp = NULL;
3312a3314
>
make: *** [check] Error 1
!!! ERROR: sys-devel/flex-2.5.4a-r5 failed.
!!! Function src_test, Line 515, Exitcode 0
!!! Make check failed. See above for details.
!!! If you need support, post the topmost build error, NOT this status message.
This has been demonstrated on x86 with both gcc-3.4 and gcc-3.5.
I noted that there are four patches used in the ebuild. Applying them on their own, I get that flex-2.5.4-glibc22.patch gives the error:
./flex -t -p ./scan.l \
| sed s,\"./scan.l\",\"scan.l\", \
| diff scan.c -
14a15
>
3312a3314
>
make: *** [check] Error 1
flex-2.5.4a-gcc3.patch gives a completed testsuite. The other gcc patch does not apply cleanly to clean sources (it depends on the first gcc patch), but when applied in this fashion, the testsuite also succeeds.
flex-2.5.4a-skel.patch gives:
./flex -t -p ./scan.l \
| sed s,\"./scan.l\",\"scan.l\", \
| diff scan.c -
1483c1483
< register char *yy_cp, *yy_bp;
---
> register char *yy_cp = NULL, *yy_bp = NULL;
make: *** [check] Error 1
This is the original error. However, both patches appear to need fixing, or if this behaviour is now expected, as it would appear from the patch, then the testsuite also needs patching.
**********
NB: There is a better testsuite in there. The make target is "bigcheck", and this passes, as it isn't quite so naive. The solution I would prefer is to use this testsuite by default instead. Fixing the ebuild is obvious so I won't bother with a patch - if someone okays me to fix this I'll commit it.
