Index | Thread | Search

From:
Stuart Henderson <stu@spacehopper.org>
Subject:
Re: GHC on OpenBSD/arm64 -> retguard fail?
To:
حبيب محمد الأمين محمد الهادي <ha.alamin@gmail.com>
Cc:
Greg Steuck <gnezdo@openbsd.org>, ports@openbsd.org
Date:
Mon, 18 Nov 2024 19:06:35 +0000

Download raw body.

Thread
On 2024/11/18 18:50, حبيب محمد الأمين محمد الهادي wrote:
> And I've already mentioned these in previous emails (with patches), but
> just to clarify, in case it helps (since I don't know what opaque FILE
> means, or how it relates to needing cross-compile scaffolding; sounds
> like an abstract FILE pointer in C):

There's a pending "flag day" ABI change in the base OS to make FILE
pointers opaque, which IIRC means that all bootstraps in ports will
need regenerating and quite likely without being able to use existing
binaries.

I don't remember how far it impacted GHC but it's conceptually not that
far from a cross-compile (i.e. compiler binaries need to be able to
produce binaries that won't run on the same system type - though it's
different versions of the same arch, rather than different archs).

It's possible there might have been some other way to workaround this
for ghc but it seemed an apt time to remind about it. Last time we did
this (for 64-bit time_t) we had to go through ghc-6.6.1 (buildable
without bootstrap), 6.10.4, 6.12.3, 7.0.4, 7.4.2. It wasn't very fun.