Re: How to drop all non-global/defined function declarations?

[Mark Wielaard <mark@klomp.org>]

Hi Dodji,

On Fri, Apr 10, 2020 at 10:48:14AM +0200, Dodji Seketeli wrote:
> Mark Wielaard <mark@klomp.org> a écrit:
> > Why are free and malloc included? They are explicitly marked as
> > undefined in the symbol table and the xml abi definition doesn't
> > include an elf-function-symbol for them?
> 
> They are included because we capture all global function and variable
> symbols, including those that are undefined.  That can be useful down
> the road for tools like abicompat:
> https://sourceware.org/libabigail/manual/abicompat.html.  It's true that
> they are not defined by the librarie so if you know that you need those,
> so yeah, maybe we should add an option to get rid of them.  Right now,
> you can't.

I have a patch to get rid of them for the dwarf reader and abidw.
Which can be extended to the xml reader and other tools if userful.

> > I think them being undefined in the symbol table make them unexported.
> > They are just there to show the library consumes that symbol.
> 
> You are right.
> 
> > That is also an abi issue, but not one I am interested in.
> 
> Well, that's the gotcha I think :-)
> 
> In your particular use case, you are not interested.  But others might
> use that information, to for instance, build the transitive closure of
> dependencies (in terms of either symbols or libraries) of the binary
> which ABI is being modelled by the current abixml file.  That's
> precisely why we are adding this information in.  As I said above, the
> abicompat tool need that information, for instance.  Users of the
> library might need it too.

You are right too. I hadn't considered abicompat. It is certainly
useful (and necessary) in the case.

> > The main issue is that they are really noise. They don't come with an
> > argument list or a real return type.
> 
> If you think having that information in is a real problem for you, we
> can find a way to let you remove it.  But I am not yet convinced that
> it's better to remove it by default because it might be useful for other
> use cases, and as it stands right now, the size of undefined entry
> points in the abixml file is quite small.
> 
> > And they make the xml abi file harder to compare (visually).
> 
> Hmmh.  I think I see your point here.  But Historically, the abixml file
> wasn't meant to be compared textually.  The tool to compare abixml files
> is abidiff and it knows how to deal with that.
> 
> > My goal here is to have an xml abi file checked into git. Then when we
> > update the abi, it should be accompanied by a change in the xml abi
> > file. The smaller and more concise we can make the xml abi file the
> > clearer it is what we are exactly changing that impacts the abi.
> 
> I get that.  We try to make this possible as much as we can.  We also
> have other constraints to satisfy.  We try to make all these ends meet.

I'll sent a patch for abidw --drop-undefined-syms which should help my case.

Cheers,

Mark