type mismatches?

[Dan Prener]

In article <MCCALPIN.91Mar19160551 at pereland.cms.udel.edu>, mccalpin at perelandra.cms.udel.edu (John D. McCalpin) writes:
|> I have been using the -qextchk option with xlf to check for type
|> mismatches at the loading step of the compilation.  Until today, I
|> have had no mismatches, so I have never seen its output.
|> 
|> So today I compile a big program and get the following message:
|> 
|> 0706-315 ERROR: Type mismatches detected by RESOLVE.
|>                  See messages under RESOLVE for more information.
|> 
|> This is *not* helpful!!!!!
|> What the hell does it mean?

I could explain the historical reasons for the message, etc., but they
wouldn't make the message any more helpful.  So I'll spare you that.

On whatever command you use to do the link (xlf, cc, xlc, or ld) add the
option

          -bloadmap:filename

where you substitute for "filename" the name of a file into which the
linker will write, alas, not a loadmap (for that you would say "map" 
instead of "loadmap") but a detailed description of what went on during
the linking process.  So, for example, if you used the command

    xlf myprog.f yourprog.f -o myprog -bloadmap:myprog.map

you could then look in the file myprog.map and find more detail about
the link process than you want to know, but among all that information
there would be a section following the word (internal linker command,
actually) "resolve" that showed the messages about type mismatches
that were referred to in the message you got.
-- 
                                   Dan Prener (prener @ ibm.com)