public inbox for binutils@sourceware.org
 help / color / mirror / Atom feed
From: Brooks Moses <bmoses@stanford.edu>
To: binutils@sources.redhat.com
Subject: Re: Using "weak externals" with ld on cygwin?
Date: Mon, 17 Apr 2006 12:52:00 -0000	[thread overview]
Message-ID: <44433349.4010005@stanford.edu> (raw)
In-Reply-To: <000001c661d4$ec0e50c0$484861cb@anykey>

Danny Smith wrote:
> Brooks Moses <bmoses at stanford dot edu> 
> wrote on: Tue, 11 Apr 2006 23:34:14 -0700 
>>My understanding of this is that I should be able to create a libA.dll
>>file with some code in it that references a function foo(), and define
>>foo() as a weak symbol aliased to bar() such that, if an executable
>>defines its own foo() or links to another dll that defines it, the libA
>>code will use that definition of foo(), but otherwise it will use bar().
>>Is that actually a correct understanding? Or am I missing something
>>somewhere?
> 
> weak symbols do not work as expected with PE DLL's.  DLL's do not allow
> undefined symbols when they are linked together.  DLLs are executables
> with their own startup entry point, their own main, their own at_exit
> table, etc. Hence a DLL built from this:
[...]
> Weak extern synbols work fine in PE  with static objects and archives,
> but I don't
> know how to do exactly what you want with dll's.  Using
> LoadLibrary/GetProcAddress API might work for you.

Thanks!  Sounds like the best answer to my "How do I do this with dlls?" 
question is "Don't; find a workaround instead" -- making my code work 
with the LoadLibrary API would probably be as much work as simply fixing 
it so I don't need the circularly-referencing dlls, and the latter is a 
much more elegant solution.

>>I've been trying, with a copy of the release version of gcc 4.1.0 and
>>the latest Cygwin version of ld (which reports GNU ld version 2.16.91
>>20050610 as its version), to write some code that will do that.
>>Everything I try seems to either give me an error like "Cannot export
>>.weak.__Z3foov.__Z3barv: symbol not found" or else always uses bar()
>>regardless of whether I supply a foo() elsewhere. 
> 
> Exporting .weak._foo._bar in the above is an ld error.  I think this
> kind of
> '.'-concatted names get interpreted as a forward export, eg the loader
> will look
> for a symbol 'weak' in another dll module with the unsual name
> '_foo._bar'. 
> This can be fixed easily enough by adding ".weak." to the list of
> prefixes that
> exclude symbols from automatic export, like so:
[...]

Should I file a bug report on this, then?  If so, where?

Thanks much!
- Brooks

  reply	other threads:[~2006-04-17  6:19 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-04-17  6:19 Danny Smith
2006-04-17 12:52 ` Brooks Moses [this message]
  -- strict thread matches above, loose matches on Subject: below --
2006-04-12  8:40 Brooks Moses

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=44433349.4010005@stanford.edu \
    --to=bmoses@stanford.edu \
    --cc=binutils@sources.redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).