head	1.3;
access;
symbols
	pkgsrc-2026Q2:1.3.0.12
	pkgsrc-2026Q2-base:1.3
	pkgsrc-2026Q1:1.3.0.10
	pkgsrc-2026Q1-base:1.3
	pkgsrc-2025Q4:1.3.0.8
	pkgsrc-2025Q4-base:1.3
	pkgsrc-2025Q3:1.3.0.6
	pkgsrc-2025Q3-base:1.3
	pkgsrc-2025Q2:1.3.0.4
	pkgsrc-2025Q2-base:1.3
	pkgsrc-2025Q1:1.3.0.2
	pkgsrc-2025Q1-base:1.3;
locks; strict;
comment	@# @;


1.3
date	2025.01.27.15.08.07;	author pho;	state Exp;
branches;
next	1.2;
commitid	6D80fPPLpq9ms8HF;

1.2
date	2025.01.12.19.06.27;	author pho;	state dead;
branches;
next	1.1;
commitid	gdBqtpho2aX0geFF;

1.1
date	2025.01.12.15.02.35;	author pho;	state Exp;
branches;
next	;
commitid	9CgagfuOmuSlUcFF;


desc
@@


1.3
log
@lang/ghc98: Replace bootkits with GHC 9.8.2

NFCI where they work. The resulting compiler produces the same binary as
before so recursive revbump is not required. This is actually a rework of
my previous changes that I had to revert.

Replacing them with 9.8.x is necessary for the upcoming GHC 9.10. I'm going
to import lang/ghc910, update every Haskell package in pkgsrc, then switch
the default compiler after that.

As a bonus this also contains a fix for one of the long-standing issues on
Darwin: GHC was failing to build any Haskell packages that contain C++
code. This affected security/hs-digest and therefore Pandoc too.

Note that GHC 9.8.2 is going to be the last version that supports
FreeBSD/i386. The upstream has apparently stopped caring about i386
platforms and I'm already seeing several signs of it bitrotting away. In
fact ghci, the REPL for GHC, has already stopped working for years, because
it immediately segfaults by exhausting the ridiculously small stack space
of i386. But I have received zero complaints so far which probably means we
can just let it go.
@
text
@$NetBSD: hadrian-splitmix.patch,v 1.1 2025/01/12 15:02:35 pho Exp $

Allow it to build with GHC 9.8.x. Already fixed in upstream.

--- splitmix.cabal.orig	2025-01-12 02:00:28.583229536 +0000
+++ splitmix.cabal
@@@@ -80,8 +80,8 @@@@ library
   -- ghc-options: -fplugin=DumpCore -fplugin-opt DumpCore:core-html
 
   build-depends:
-      base     >=4.3     && <4.19
-    , deepseq  >=1.3.0.0 && <1.5
+      base
+    , deepseq
 
   if flag(optimised-mixer)
     cpp-options: -DOPTIMISED_MIX32=1
@


1.2
log
@lang/ghc98: Fix build by reverting changes I made today

Turned out I committed these changes prematurely. I could only do it after
rebuilding all of the bootkits.
@
text
@@


1.1
log
@lang/ghc98: Replace the bootkit for NetBSD/aarch64 with GHC 9.8.2

NFCI. It produces the same binary as before so revbump is not required.

This is necessary for the upcoming GHC 9.10. I'm going to replace the
bootkit for each platform one by one, import lang/ghc910, update every
pkgsrc package written in Haskell, then switch the default compiler after
that.

I might have to give up on FreeBSD/i386 though. The upstream apparently
stopped caring about i386 platforms and I'm already seeing several signs of
it bitrotting away. In fact ghci, the REPL for GHC, has already stopped
working for years but I have received zero complaints so far which probably
means we can just let it go.
@
text
@d1 1
a1 1
$NetBSD$
@