[Gm2] Re: debugging Modula-2

[John O Goyo]

Gaius Mulley wrote:
> Hi John,
>
> I've just had to fix gdb so that it can debug
> gm2/gm2-libs-iso/WholeConf.mod, in particular gdb attempted to go into an
> infinite recursive loop when resolving:
>
>
>   ScanClass =  (* Values of this type are used to classify input to
>   finite state scanners *)
>   (
>     padding,   (* a leading or padding character at this point in the
>   scan - ignore it *)
>     valid,     (* a valid character at this point in the scan - accept
>   it *)
>     invalid,   (* an invalid character at this point in the scan -
>   reject it *)
>     terminator (* a terminating character at this point in the scan
>   (not part of token) *)
>   );
>
>   ScanState =  (* The type of lexical scanning control procedures *)
>     PROCEDURE (CHAR, VAR ScanClass, VAR ScanState);
>
> the procedure type.  As the above declaration occurs in ConvTypes in
> the ISO libraries - any ISO program would eventually be non debug-able.
>
> So my question is whether this is the same bug you have seen on Solaris -
> albeit you are using dbx (and stabs?).  But are you able to debug non
> iso programs using non iso libraries?

No, the problem on Solaris 10 was that, at first, gdb considered a debugging
binary to be corrupt but dbx could handle it.  Later on, both gdb and dbx
considered such a binary corrupt and refused to load it.  I did a bit of
investigation and found the cause for the first woe to be multiple dwarf
sections.  Dbx handled that but gdb declared it corrupt.  (A brief look at the
dwarf standard seemed to indicate multiple dwarf sections in that context
legal but that is neither here nor there.  I did find a note in the gdb-mail
archives of a similar problem in the MIPS version that was fixed.)  In later
releases of gm2, both debuggersdeclared debugging binaries to be corrupt and
refused to load them.  This applied to non-ISO binaries as well as ISO
binaries.

john