This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB 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, testsuite] Clean up gdb.trace results


On 2018-10-10 22:01, Sandra Loosemore wrote:
Thanks, I tried that and made a couple tweaks to the patch to fix
those errors.  Is this version OK?

-Sandra

Hi Sandra,

I'd just like to understand the situation a bit more. From what I understand, trace-common.h is only useful for fast tracepoint tests. Let's take the actions.exp test, for example. It uses actions.c, which includes trace-common.h, therefore you added a gdb_trace_common_supports_arch check. However, actions.exp does not rely on fast tracepoint support at all. actions.c happens to be used by another test that tests fast tracepoints. The result is that if an architecture supports tracepoints, but not fast tracepoints, the actions.exp test will be skipped even though it would be relevant.

Trying to read between the lines: if your target did not support tracepoints at all, then the gdb_target_supports_trace calls would be enough. If your target did support regular and fast tracepoints, then you would just add support for it in trace-common.h, because you would want to run the fast tracepoint tests. So the remaining combination is that your target supports regular tracepoints, but not fast tracepoints. Is it the case?

If so, I think the right fix would be to untangle fast tracepoints tests from regular tracepoints tests. The goal being to make it possible to run as many tests as possible against targets that only support regular tracepoints.

Or maybe I'm completely lost, in which case can you clarify what your use case is?

Thanks,

Simon


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