Debugging support - Re: g77 features lacking in gfortran - PR19292

Debugging support - Re: g77 features lacking in gfortran - PR19292

[Wu Zhou]

Hi Thomas,

On Sun, 23 Oct 2005, Thomas Koenig wrote:

> On Sat, Oct 22, 2005 at 08:14:50AM +0200, Paul Thomas wrote:
> 
> > >22244 dimension information is lost for multi-dimension array - 
> > >primarily gdb and optimization problem.
> 
> I think that one's fairly important.  I wouldn't call debugging
> support OK if you can't look at arrays properly.  Unfortunately,
> I don't even have the first idea to where I should start looking to
> fix this...

GDB also has some other problems while working with gfortran, such as the 
printing of one-dimensional array, function evaluation, handling of common 
block variable and so on...

Some of these problems stems from the DWARF output of gfortran, some lies 
in the code of gdb, some might involve both.  I has been working on some 
of these problem at gdb's side.  I am now thinking of whether there are 
any specific requirement from gfortran community for GDB.  Did any of you
have a big picture about how well gdb work with gfortran?  Did you 
have a list of things that want GDB to support first?  Part of my effort 
is to extend gdb's testsuite with reference to gfortran compiler.  So I am 
very happy to receive any feedback on these.  Thanks a lot in advance.

Best Regards
- Wu Zhou  

Re: Debugging support - Re: g77 features lacking in gfortran - PR19292

[Thomas Koenig]

On Tue, Oct 25, 2005 at 02:18:33PM +0800, Wu Zhou wrote:

> GDB also has some other problems while working with gfortran, such as the 
> printing of one-dimensional array, function evaluation, handling of common 
> block variable and so on...

... variables from modules (I've just reported this as PR 24526),
allocatable arrays (PR 17905) and so on.

> 
> Some of these problems stems from the DWARF output of gfortran, some lies 
> in the code of gdb, some might involve both.  I has been working on some 
> of these problem at gdb's side.  I am now thinking of whether there are 
> any specific requirement from gfortran community for GDB.  Did any of you
> have a big picture about how well gdb work with gfortran?  Did you 
> have a list of things that want GDB to support first?

My (personal) vote would go to supporting arrays (one- and
multidimensional) (PR 22244) should be pretty hight on the list,
followed by variables in common blocks (PR 23057).

Any help you can provide is certainly very welcome!

	Thomas

Re: Debugging support - Re: g77 features lacking in gfortran - PR19292

[Wu Zhou]

Hi Thomas,

On Tue, 25 Oct 2005, Thomas Koenig wrote:

> On Tue, Oct 25, 2005 at 02:18:33PM +0800, Wu Zhou wrote:
> 
> > GDB also has some other problems while working with gfortran, such as the 
> > printing of one-dimensional array, function evaluation, handling of common 
> > block variable and so on...
> 
> ... variables from modules (I've just reported this as PR 24526),
> allocatable arrays (PR 17905) and so on.

PR 24526 could be worked around by using __foo__a, where foo is the 
modular name and a being the variable name.  To handle a directly, maybe 
some changes in gdb is needed.  I had added a comment in the bug report.

For 17905, if gfortran could output similar debuginfo as IFC does, gdb 
could handle that.  I ever pointed that out in the bug report.

BTW, if you can add me into the cc-list when you find a bug is with 
reference to debugging, that will be highly appreciated.

> > 
> > Some of these problems stems from the DWARF output of gfortran, some lies 
> > in the code of gdb, some might involve both.  I has been working on some 
> > of these problem at gdb's side.  I am now thinking of whether there are 
> > any specific requirement from gfortran community for GDB.  Did any of you
> > have a big picture about how well gdb work with gfortran?  Did you 
> > have a list of things that want GDB to support first?
> 
> My (personal) vote would go to supporting arrays (one- and
> multidimensional) (PR 22244) should be pretty hight on the list,
> followed by variables in common blocks (PR 23057).

IIRC, Andrew Pinski ever said that Paul Brook is looking at the way to fix 
PR 22244, but I didn't see any update for a long time.  Don't know if Paul 
ever had any time on this?  Maybe you can contact him for clue on how to 
fix that.  If you can work out a way in gfortran to correctly represent 
arrary dimension information in DWARF output, I am very happy to test 
that.  For PR 23057, if gfortran could output standard compliant DWARF 
output, gdb (other debugger as well IMHO) could handle that.  But maybe I 
can take a look at gdb code to see whether it could be resolved in gdb's 
side.   

> Any help you can provide is certainly very welcome!

Great pleasure.  I am more than happy to contribute to that.  :-)

Best Regards
- wu Zhou

Re: Debugging support - Re: g77 features lacking in gfortran - PR19292

[Thomas Koenig]

I have created PR 24546 as a meta-bug for tracking debugging problems
in gfortran.  Feel free to add any bugs that I may have missed.

	Thomas

Re: Debugging support - Re: g77 features lacking in gfortran - PR19292

[Wu Zhou]

It is quite complete as I see.  Will add into this meta-bug if I find any 
new debugging problems.

On Wed, 26 Oct 2005, Thomas Koenig wrote:

> I have created PR 24546 as a meta-bug for tracking debugging problems
> in gfortran.  Feel free to add any bugs that I may have missed.