This is the mail archive of the cygwin-xfree@cygwin.com mailing list for the Cygwin XFree86 project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Esound 0.2.8 is dead - Esound 0.2.23 compiles fine


On Mon, 2001-10-29 at 16:07, Christopher Faylor wrote:
> On Mon, Oct 29, 2001 at 03:55:26PM +1100, Robert Collins wrote:
> >On Mon, 2001-10-29 at 15:00, Christopher Faylor wrote:
> >>Ah.  I was wondering why errors weren't bein displayed already.  Now I
> >>understand the distinction.
> >>
> >>If we added a "--warn-bad-auto-export" as an ld option for building
> >>DLLs, could we flag errors that way?  Would it give false negatives if
> >>we warned about multi-word exports?
> >
> >Thats what I read into Chuck's "anything based on the base address of
> >the next export is bound to be flakey".
> 
> I read that as saying that it wasn't bad for a "normal" export but it
> was fatal when used with auto-import.  So, you wouldn't want to issue
> a warning unless you were interested in using the DLL as an auto-import
> DLL.
> 
> Or do I have it backwards?

I think we have two facets of the same thing.
1) Exported multi-word variables are normal, and only an issue for the
current auto-import code. The linker may be able to discriminate between
explicitly export and auto exported variables, and only warn on auto
exported variables - which is likely to be quite precise as undecorated
build headers implies that the user headers will also be undecorated.
2) The detect of multi-word variables by offset deltas would be fraught
with false positives.

Rob


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]