Thiago,
This is a false alarm.. Sorry about that.
This has nothing to do with gdb, it complains about a spec2006 benchmark. not even a gcc testsuite issue. It is just a benchmark instability build issue.
tonto benchmark was unstable due to an issue when parallelizing make, with remote compilation (gfortran), and local changes at the same time (preprocessing). A fix is already used for spec2017, and proposed the same fix for spec2006 : - https://review.linaro.org/c/toolchain/jenkins-scripts/+/42996 - https://review.linaro.org/c/toolchain/bmk-scripts/+/42997
For details, the compilation error trace can be seen here : bkp-01.tcwglab:/home/tcwg-benchmark/results-tx1_64/tcwg_bmk-code_speed-spec2k6/build-gnu-aarch64-master-O3/24149/tcwg-bmk-tx1-04.tcwglab/CPU2006.serial.log.debug.build.1
1674108293.46: /home/tcwg-benchmark/tcwg-bmk-tx1-04.tcwglab/cpu2006-tx1_64-1/bin/ssh-serial-gfortran -O3 -fdump-statistics-asmname -fdump-tree-vect-details -save-temps=obj -Qn -c -o coppensbasisvec.fppized.o coppensbasisvec.fppized.f90 1674108293.56: coppensbasisvec.fppized.f90:985:10:
1674108293.56: 985 | type(c 1674108293.56: | 1 1674108293.56: Error: Malformed type-spec at (1) 1674108293.56: f951: Error: Unexpected end of file in 'coppensbasisvec.fppized.f90' 1674108293.56: specmake: *** [coppensbasisvec.fppized.o] Error 1 1674108293.56: specmake: *** Waiting for unfinished jobs.... 1674108298.79: Tee child 1483 exited with code 512 (exit code=2, signal=0)
It failed when building, then, when bisecting it failed sometimes, and passes some others.. which happens to finally find one commit. I would say randomly.
Regards
Le 20/01/2023 à 23:53, Thiago Jung Bauermann a écrit :
ci_notify@linaro.org writes:
After gdb-13-branchpoint-527-g46758593515 commit 4675859351582f017b495ff13fb2ea72a99834af Author: Tom Tromeytom@tromey.com
Rename to allow_ifunc_testsThis is a GDB commit so I had a look at this failure. The commit can't be responsible for it though, because it only changes some .exp files in the GDB testsuite.
the following benchmarks slowed down by more than 3%:
- 465.tonto failed to build
I couldn't find the actual build failure in the job artifacts, so I ran the instructions for reproducing the failure on my dev container in tcwg-jade-03. It did fail, but with a slightly different error:
416.gamess failed to build.
I suppose this is GCC's test gcc/testsuite/gfortran.dg/fmt_read_5.f, so I went to directory abe/builds/aarch64-unknown-linux-gnu/aarch64-linux-gnu/gcc-gcc.git~master-stage2/gcc and ran:
$ make check-fortran RUNTESTFLAGS=dg.exp=fmt_read_5.f
Which passed, so I'm stuck now.
Regarding the failure of 465.tonto in the original CI report, I couldn't find any GCC test that could be related to it.