skip "attach" tests when testing against stub-like targets (was: Re: [PATCH 5/5] Test attaching to a program that constantly spawns short-lived threads)

From: Pedro Alves <palves@redhat.com>

  On 01/05/2015 07:01 PM, Breazeal, Don wrote:
> On 12/17/2014 4:02 PM, Pedro Alves wrote:
> 
>> diff --git a/gdb/testsuite/lib/gdb.exp b/gdb/testsuite/lib/gdb.exp
>> index 08087f2..83fa1d0 100644
>> --- a/gdb/testsuite/lib/gdb.exp
>> +++ b/gdb/testsuite/lib/gdb.exp
>> @@ -3413,12 +3413,36 @@ proc gdb_exit { } {
>>      catch default_gdb_exit
>>  }
>>
>> +# Return true if we can spawn a program on the target and attach to
>> +# it.
>> +
>> +proc can_spawn_for_attach { } {
>> +    # We use TCL's exec to get the inferior's pid.
>> +    if [is_remote target] then {
>> +	return 0
>> +    }
>> +
>> +    # The "attach" command doesn't make sense when the target is
>> +    # stub-like, where GDB finds the program already started on
>> +    # initial connection.
>> +    if {[target_info exists use_gdb_stub]} {
>> +	return 0
>> +    }
>> +
>> +    # Assume yes.
>> +    return 1
>> +}
>> +
> This solves the problem that I was working on here:
> 
> https://sourceware.org/ml/gdb-patches/2014-12/msg00520.html
> 
> When I call can_spawn_for_attach in the misbehaving attach tests I was
> working on, they no longer spawn processes for 'target remote' that they
> can't attach to.  Thanks!

Great!

> 
>>  # Start a set of programs running and then wait for a bit, to be sure
>>  # that they can be attached to.  Return a list of the processes' PIDs.
>>
>>  proc spawn_wait_for_attach { executable_list } {
>>      set pid_list {}
>>
>> +    if ![can_spawn_for_attach] {
>> +	error "can't spawn for attach with this target/board"
>> +    }
> 
> Should this be calling "error", or should it call something like
> "untested" or "unsupported", since it isn't expected to work in these cases?

The idea is that all .exp files that use spawn_wait_for_attach
would have already checked can_spawn_for_attach early, and skipped the
tests if false.  That makes is a test bug to see a call to
spawn_wait_for_attach if can_spawn_for_attach is false.

I should have really split those hunks out to a separate patch and
added calls to can_spawn_for_attach in all tests that are using
spawn_wait_for_attach already.  Like below.  WDYT?

(There are probably other attach tests that don't use
spawn_wait_for_attach that need the can_spawn_for_attach too.
We can do this incrementally.)

--------
From a7d938a9ca3762ea195a20b796772865a47283b6 Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Wed, 7 Jan 2015 15:38:02 +0000
Subject: [PATCH] skip "attach" tests when testing against stub-like targets

We already skip "attach" tests if the target board is remote, in
dejagnu's sense, as we use TCL's exec to spawn the program on the
build machine.  We should also skip these tests if testing with
"target remote" or other stub-like targets where "attach" doesn't make
sense.

Add a helper procedure that centralizes the checks a test that needs
to spawn a program for testing "attach" and make all test files that
use spawn_wait_for_attach check it.

gdb/testsuite/
2014-12-17  Pedro Alves  <palves@redhat.com>

        * lib/gdb.exp (can_spawn_for_attach): New procedure.
        (spawn_wait_for_attach): Error out if can_spawn_for_attach returns
        false.
	* gdb.base/attach.exp: Use can_spawn_for_attach instead of
	checking whether the target board is remote.
	* gdb.multi/multi-attach.exp: Likewise.
	* gdb.python/py-sync-interp.exp: Likewise.
	* gdb.server/ext-attach.exp: Likewise.
	* gdb.python/py-prompt.exp: Use can_spawn_for_attach before the
	tests that need to attach, instead of checking whether the target
	board is remote at the top of the file.
---
 gdb/testsuite/gdb.base/attach.exp           |  3 +--
 gdb/testsuite/gdb.base/solib-overlap.exp    |  3 +--
 gdb/testsuite/gdb.multi/multi-attach.exp    |  3 +--
 gdb/testsuite/gdb.python/py-prompt.exp      |  9 ++++-----
 gdb/testsuite/gdb.python/py-sync-interp.exp |  3 +--
 gdb/testsuite/gdb.server/ext-attach.exp     |  3 +--
 gdb/testsuite/lib/gdb.exp                   | 27 +++++++++++++++++++++++++++
 7 files changed, 36 insertions(+), 15 deletions(-)

skip "attach" tests when testing against stub-like targets (was: Re: [PATCH 5/5] Test attaching to a program that constantly spawns short-lived threads)

Commit Message

Comments

Patch