This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch] Disable testcases with too large bit-fields.


Kazu Hirata wrote:
Attached is a patch to disable testcases with too large bit-fields.

There is no analysis here explaining why this is the right thing to do. The execute/comp-goto-1.c patch is particularly interesting, because it already has a check for INT_MAX == 32767, which implies that this used to work for 16 bit targets in the past. The patch that made it work is here:
http://gcc.gnu.org/ml/gcc-patches/2000-12/msg01042.html
So why doesn't it work now? Did something accidentally break? Or did someone intentionally change semantics? If something accidentally broke, then you would hiding the real failure by changing the testsuite. So it is important to know what is going on here.


The answer is the latter. A patch changed the "width of '%s' exceeeds its type" diagnostic from a pedwarn to an error, causing the testsuite failures. This was an intentional change.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=3347
It helps the patch reviews if you include info like this, instead of making us look it up ourselves.


So we have to fix the testsuite. We could fix this problem by changing unsigned and unsigned int to unsigned long, but there is no guarantee that the modified tests will still be testing the same problem they were intended to test. Disabling the tests for 16-bit targets if fine with me.

This is OK.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com


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