Re: [PATCH v5 2/2] PR 20569, segv in follow_exec

[Pedro Alves <palves at redhat dot com>]

On 10/26/2016 05:04 PM, Luis Machado wrote:
> On 10/26/2016 09:28 AM, Luis Machado wrote:
>> On 10/25/2016 06:39 PM, Pedro Alves wrote:

>>> +# This test exercises PR20569.  GDB would crash when attempting to
>>> follow
>>> +# an exec call when it could not resolve the path to the symbol file.
>>> +# This was the case when an invalid sysroot is provided.
>>> +
>>> +standard_testfile foll-exec.c
>>> +
>>> +global binfile
>>> +set binfile [standard_output_file "foll-exec"]
> 

[FYI, I pushed this in already yesterday.]

> On further inspection, i think this violates the testcase cookbook's
> rules? It is reusing the foll-exec.c sources, but it should compile to a
> unique executable name.
> 
> On the other hand, foll-exec.c relies on its own name to do the trick.
> It replaces foll-exec with execd-prog.
> 
> Is this still a problem now that we have unique directories for each
> testcase?

Yeah, I think it's probably not such a big deal nowadays as it used to be.

The case where it'd maybe still help a little bit is the remote host/target
case.  In that case, we'll still download the files to a single remote
directory.  We can't do parallel testing in that case, so the downside in
general is that the overwriting may make it a bit harder to debug failing
testcases after the fact.  Maybe not that much of a problem, since we have
the separate copies on the host side.  We already have this overwriting
on the target side in case we have tests with the same name in
different gdb.foo/ subdirectories, I think:

$ ls testsuite/gdb.*/*.exp | sed 's,testsuite/gdb.[a-z]*/,,g' | sort | uniq -c | sort -nr | head -n 30
      3 types.exp
      3 print.exp
      3 gcore.exp
      3 exprs.exp
      2 vla-datatypes.exp
      2 start.exp
      2 solib-symbol.exp
      2 solib.exp
      2 sizeof.exp
      2 range-stepping.exp
      2 ptype.exp
      2 pr11022.exp
      2 pending.exp
      2 methods.exp
      2 maint.exp
      2 logical.exp
      2 integers.exp
      2 hello.exp
      2 exception.exp
      2 demangle.exp
      2 debug-expr.exp
      2 data.exp
      2 complex.exp
      2 charset.exp
      2 callfuncs.exp
      2 break.exp
      2 backtrace.exp
      2 arrayidx.exp
      2 annota3.exp
      1 xfullpath.exp

Thanks,
Pedro Alves