Re: [PATCH 1/2] Test case for entry values.

[Vidya Praveen <vidyapraveen@arm.com>]

On Wed, Sep 11, 2013 at 12:43:24AM +0100, Yao Qi wrote:
> On 09/10/2013 11:30 PM, Vidya Praveen wrote:
> > OK. But isn't it better to have the condition (!gdb_target_supports_trace)
> > that checks if the target supports tracing, in the beginning of the test
> > rather than much later?
> 
> This part of test is about testing entry values, and the bottom part 
> (added by patch 2/2) is about testing unavailable entry values when 
> examining trace frames.  This part is not related to tracing, so we 
> can't use gdb_target_supports_trace to check.
> 
> See my description in "PATCH 0/2"
> 
> > Patch 1/2 is to generate dwarf using Dwarf Assembler to test "entry values"
> > are shown correctly.  At this point, gdb.trace/entry-values.exp is still
> > a dwarf test, nothing to do with trace.  Patch 2/2 is to use tracepoint,
> > to collect data, to test what happen when argument is available and entry
> > value is not.
> 
> https://sourceware.org/ml/gdb-patches/2013-08/msg00327.html

However, it scans for the 'call' instruction regardless of the target. Though
your comment explains it, the test is still target specific. It should check
for the target before scanning for 'call' and not assume the $sp for all 
targets except for those you already check.

Regards
VP