Re: [6/6] Add a partial COFF library support to the winedump

[Eric Pouech]

Dmitry Timoshkov a écrit :
> Hello,
>
> this implementation is able to parse imports libraries from the PSDK
> and hopefully should help interested parties to check/synchronize ordinal
> imports between Wine and Windows.
>
> The dumper is not able (yet) to parse import libraries with the long format
> described in Portable Executable and Common Object File Format Specification
> and uuid.lib.

I think it's a bad idea to "force" the kind of file on the command line 
for dumping
I'd rather see winedump guess the file format based on the file header, 
as we do for .exe, .dll, .mdmp, .pdb, .dbg...
we didn't do it for emf and lnk because it seems they don't have a 
simple header to find it out, but .lib files definitively have

you should also provide support for the -C option (symbol demangling)

BTW, I have a couple of patches for debug files dumping (.pdb, .dbg...), 
that should help the dumping of debug information from a coff file (when 
needed)

A+

[6/6] Add a partial COFF library support to the winedump

[Dmitry Timoshkov]

"Eric Pouech"  wrote:
> I think it's a bad idea to "force" the kind of file on the command line 
> for dumping
> I'd rather see winedump guess the file format based on the file header, 
> as we do for .exe, .dll, .mdmp, .pdb, .dbg...

I thought about that, but that's definitely could be done as a separate
work.
> we didn't do it for emf and lnk because it seems they don't have a 
> simple header to find it out, but .lib files definitively have

Both emf and lnk kind of files do have a distinct headers: emf files
have ENHMETAHEADER in EMR_HEADER record with all kinds of recognizeable
information including a signature, and lnk files have LINK_HEADER with
recognizeable dwSize and MagicGuid fields.
> you should also provide support for the -C option (symbol demangling)

I haven't needed it at this point and most PSDK import libs don't need it,
but for C++ import libraries it would be nice to have.
> BTW, I have a couple of patches for debug files dumping (.pdb, .dbg...), 
> that should help the dumping of debug information from a coff file (when 
> needed)

Great! And many thanks for the review!

[6/6] Add a partial COFF library support to the winedump

[Dmitry Timoshkov]

"Dmitry Timoshkov"  wrote:
> > I think it's a bad idea to "force" the kind of file on the command line 
> > for dumping
> > I'd rather see winedump guess the file format based on the file header, 
> > as we do for .exe, .dll, .mdmp, .pdb, .dbg...
>
> I thought about that, but that's definitely could be done as a separate
> work.

I'd go even further by introducing is_lnk(const char *fname) and similar
is_emf(), is_pe(), is_le() functions as a general recognition interface
in the option_table[]. Also I'd get rid the "dump" command line switch
altogether and made the DUMP the default mode.

[6/6] Add a partial COFF library support to the winedump

[Eric Pouech]

Dmitry Timoshkov a écrit :
> "Eric Pouech"  wrote:
> > I think it's a bad idea to "force" the kind of file on the command 
> > line for dumping
> > I'd rather see winedump guess the file format based on the file 
> > header, as we do for .exe, .dll, .mdmp, .pdb, .dbg...
>
> I thought about that, but that's definitely could be done as a separate
> work.

as we agree on the goal, patches that are not the tracks to that should 
be changed IMO
> > you should also provide support for the -C option (symbol demangling)
>
> I haven't needed it at this point and most PSDK import libs don't need 
> it,
> but for C++ import libraries it would be nice to have.

the demangling code is already in place, so that's easy to do (note that 
the current demangling code in winedump lacks quite a few enhancements 
that are present in msvcrt)
A+