This is the mail archive of the
mailing list for the Cygwin project.
Re: unison-2.48 build fails with latest ocaml and flexdll (attn: Achim)
- From: Achim Gratz <Stromeko at nexgo dot de>
- To: cygwin-apps at cygwin dot com
- Date: Mon, 03 Apr 2017 08:23:18 +0200
- Subject: Re: unison-2.48 build fails with latest ocaml and flexdll (attn: Achim)
- Authentication-results: sourceware.org; auth=none
- References: <h1hvhblmsho84o6ofuetks19agfo3u9ro4-e09XROEemail@example.com> <firstname.lastname@example.org> <email@example.com> <41t6ibtbcojbmar8blrc80fk8tmplbbuj3-e09XROEfirstname.lastname@example.org> <5723B6C7.email@example.com> <firstname.lastname@example.org> <email@example.com> <5mt0bcpirrh22jobp0qtm5q0hsfqqncjt8-e09XROEfirstname.lastname@example.org> <email@example.com> <firstname.lastname@example.org> <email@example.com>
Yaakov Selkowitz writes:
> Achim, I think /usr/lib/ocaml needs to be ignored by autorebase;
> flexlink'ed DLLs are loaded very differently, and rebasing seems to
> break them.
I will be on a business trip for the rest of the week. I can look at
excepting them, but I'd rather introduce a more general mechanism for
such things, I'll have to think about that.
The question is however if these actually work when they're not rebased
while some other library occupies their address space. I suspect that'd
fail unless they are ASLR enabled. Plus we need to make sure they don't
slip into the rebase space and block another Cygwin DLL trying to load
later. Can somebody please test that?
While on the topic of ASLR, I've been experimenting with generally
making all (or most) Cygwin DLL ASLR enabled and avoid rebasing
altogether (that was triggered by some 32bit machine that I couldn't
switch over to a 3GB user VM for some reason). It seems like that
should work, although it's unclear which subset of libraries must load
to fixed addresses (rebased or not). Also, I'm not quite certain which
address space ASLR will use and if there's a "safe" space that it will
never occupy. Unfortunately I haven't had much time to spend on that,
so if anybody feels inclined to try it I'll see if I can put something
into autorebase to support it.
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+
Waldorf MIDI Implementation & additional documentation: