Fix that rootcling magically passes with 256 errors #898
Conversation
|
Starting build on |
core/dictgen/src/rootcling_impl.cxx
Outdated
| // FIXME: We shouldn't count errors at all via the rootcling return code. See the warning about this from the | ||
| // libc documentation: http://www.gnu.org/savannah-checkouts/gnu/libc/manual/html_node/Exit-Status.html | ||
| // This probably requires adapting roottest too as we check the exit code values there. | ||
| auto mainRetVal = nerrors + retVal; |
There was a problem hiding this comment.
I totally agree. I am not aware of roottest checking exit code values, except for "0 or not 0". Can you please set mainRetVal to 0 or 1, and see what roottest says about that?
There was a problem hiding this comment.
I can even point you to an example line where this breaks: https://github.com/root-project/roottest/blob/master/root/meta/genreflex/CMakeLists.txt#L88 :)
But let's indeed remove this error counting mechanic and follow libc recommendations.
There was a problem hiding this comment.
I stand corrected, but now I have no idea how we passed those tests :)
Teemperor
force-pushed
the
FixRootclingSucceedOn256Errors
branch
from
eb52d4e to
02d5d86
Compare
|
Starting build on |
Teemperor
force-pushed
the
FixRootclingSucceedOn256Errors
branch
from
02d5d86 to
a4fef1d
Compare
|
Starting build on |
We currently have a byte overflow in the exit code for rootcling, so with N*256 errors we actually return 0 from our process. This patch removes this functionality and rootcling now only returns 0 and 1.
|
Just for the record, all the Jenkins builds passed, but I just rewrote the commit message so Jenkins is rebuilding it without any changes. Any objections to get this merged? |
|
Do it! |
|
Build failed on slc6/gcc49. Errors:
|
3 participants
No description provided.