This is the mail archive of the
cygwin
mailing list for the Cygwin project.
Re: Question of the necessity of rebaseall
On Wed, May 13, 2009 at 10:06 PM, Eric Blake wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> According to Lenik on 5/13/2009 7:49 PM:
>>> You have it backwards. Forking doesn't break the relocation. Relocation
>>> breaks forking. cygwin1.dll needs to have a very special memory layout to
>>> implement the fork semantics in Win32. If this memory layout is
>>> disrupted, fork breaks.
>>>
>> Could you explain in more detail? I can't find any document about this
>> special memory layout.
>
> Read the source. ÂThis link is a bit old, but still captures the essence:
> http://cygwin.com/cgi-bin/cvsweb.cgi/src/winsup/cygwin/how-cygheap-works.txt?rev=1.5&content-type=text/x-cvsweb-markup&cvsroot=src
>
> Remember, the semantics of fork is that BOTH processes (the parent and
> child) must see the SAME memory, and that includes all shared libraries
> being mapped at the SAME location. ÂBut since Windows doesn't provide a
> native fork, the child must remap everything that the parent had, and hope
> that it lands at the same place. ÂRebasing improves the chance that the
> child will remap, because there are fewer dlls to be remapped in an
> arbitrary order.
Is this a place where using vfork() instead of fork() helps (where
it's applicable, of course)? If so, we might be able to reduce the
number of rebase failures in the future just by trying to push other
projects to use vfork wherever it's substitutable for fork...
~Matt
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/