Daniel Jacobowitz wrote:
> On Mon, Oct 13, 2008 at 01:38:36PM +0200, Denis PILAT wrote:
>   
>> But the userdef.exp testsuite only expects gdb to return the 2nd  
>> semantic, which is fine with some gdb based debuggers, but not for all.
>>     
>
> What does "some gdb based debuggers" mean?  The testsuite is supposed
> to accomodate the version of GDB it's shipped with, not other
> versions.
>   
I would have talked about compiler generated information. I have a 
compiler that generates debug information so that gdb prints this class 
as a class, and not a struct!
The target is not a native one but an internal ST one, but I guess that 
there are other targeted gdbs that behave the same, and this behavior is 
correct to me! My patch was just to let some gdbs avoid an error in the 
testsuite that is not a real error, unless there is a consensus that 
tells simple class must be represented as structures.

>   
>> @@ -153,7 +153,14 @@
>> gdb_test "print c" "\\\$\[0-9\]* = {m = {z = .*}}"
>>     
>
> Also, this is whitespace-damaged.
>   
Sorry, I don't see it, I re-attach my patch as a real attachment, may be 
my mailer did something wrong

-- 
Denis