[cdt-dev] RE: Something for DSF GDB feature-parity with CDI GDB?

[Marc Khouzam <marc.khouzam@xxxxxxxxxxxx>]

 > -----Original Message-----
> From: cdt-dev-bounces@xxxxxxxxxxx 
> [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Elmenthaler, Jens
> Sent: Tuesday, March 30, 2010 11:26 AM
> To: CDT General developers list.
> Subject: [cdt-dev] Something for DSF GDB feature-parity with CDI GDB?
> 
> I just figured out that the CDI GDB does not suffer (just by 
> fortune, I guess) a bug in gdb 7.x, that the DSF GDB runs 
> into: http://sourceware.org/bugzilla/show_bug.cgi?id=11407
> 
> In the worst - but not quite unusual case - the variables 
> view remains empty for the entire duration of the function or 
> method you debug, if
> 
> -          You use a gdb 7.0 later
> 
> -          And have pretty printers for some local variables

Does this only happen when you are testing the new Pretty-printer
feature you are adding to DSF-GDB, or does it happen even without
that new code?

> The difference between CDI GDB and DSF GDB simply is that CDI 
> GDB does not print the values with -stack-list-locals. I had 
> a quick look and tested it in my C++ debug session and think 
> DSF GDB could possibly also live without printing values in 
> -stack-list-locals. I couldn't see any usage of the value.

If this is really a problem, it should not be a problem to fix it.
In fact, with the commandFactory, we could make the change
only for GDB 7.0 and later (until there is a fixed version of gdb)

> My personal impression after working a couple of days with 
> DSF GDB and gdb 7.x plus STL pretty printers is that I should 
> not switch from CDI to DSF.

This is the kind of issues that need to be reported to properly
improve DSF-GDB.  Please elaborate on any other problems you have
encountered.  Without this information, the issues that bother
you cannot be fixed.

In this particular case, pretty-printers is a new functionality
of GDB and CDI was never enhanced to support it (and probably
never will).  The fact that it seems to work better is by pure
luck.  You have just as much chance of finding some other new
feature that will instead break CDI.  The difference is that 
we are actively working at fixing DSF-GDB when such issues are 
reported.

So please continue reporting any issues you find, and we'll be able
to properly support the many new GDB features that you won't get
with CDI (non-stop, tracepoints, multi-exec, checkpoints, etc).

> So my question, is this a bugzilla for the feature-parity 
> with CDI? I happily enter one.

If it happens currently with HEAD, then yes it is, even if it
is by fluke.
Please open the bug and we can get it fixed.

Thanks for reporting the issue.

Marc