msvcrt is only one of the libcs in MinGW. We therefore
replace explictly testing for msvcrt with the isMinGW
predicate. This lays the foundation for ucrt64 support.
This commit prepares libiconvReal to replace darwin.libiconv, allowing
it to be used with binary derivations that patch out references to the
system libiconv with one from nixpkgs.
Apple’s libiconv is based on GNU libiconv 1.11 (the last version before
it switched to LGPLv3+). Any newer releases by Apple appear to be build
system tweaks. The core sources are barely updated. This means that
Darwin users won’t get any fixes from upstream updates, and maintaining
darwin.libiconv requires dealing with a separate and different build
system (because Apple now builds it with Xcode). Fortunately, it is
possible to build upstream libiconv in a way that is compatible with
Apple’s distribution of it.
There are two things that need to happen to produce an ABI-compatible
build of libiconv:
* Existing symbols need to be exported with the `iconv_` prefix instead
of the `libiconv_` prefix. New symbols can have the `libiconv` prefix,
and one symbol in Apple’s distribution does, but older ones must have
the older prefix; and
* Reexport `libcharset.dylib` from `libiconv.dylib`. This is explained
by Apple as the result of a bug in their transition to an Xcode-based
build system.
Both these these are doable and have been done by this commit. I have
tested it with building GHC, which downloads a binary distribution as
part of its bootstrap and replaces references to the system libiconv
with darwin.libiconv. Using this patch, libiconvReal is able to work
without any changes to the GHC derivation.
Note that this patch does not actually deprecate or remove
darwin.libiconv yet. That will be done in a future patch after Darwin
support is added for aliases and deprecating packages in the `darwin`
attrset.
We can use use `stdenv.hostPlatform.isStatic` instead, and move the
logic per package. The least opionated benefit of this is that it makes
it much easier to replace packages with modified ones, as there is no
longer any issue of overlay order.
CC @FRidh @matthewbauer
I hate the thing too even though I made it, and rather just get rid of
it. But we can't do that yet. In the meantime, this brings us more
inline with autoconf and will make it slightly easier for me to write a
pkg-config wrapper, which we need.
(My OCD kicked in today...)
Remove repeated package names, capitalize first word, remove trailing
periods and move overlong descriptions to longDescription.
I also simplified some descriptions as well, when they were particularly
long or technical, often based on Arch Linux' package descriptions.
I've tried to stay away from generated expressions (and I think I
succeeded).
Some specifics worth mentioning:
* cron, has "Vixie Cron" in its description. The "Vixie" part is not
mentioned anywhere else. I kept it in a parenthesis at the end of the
description.
* ctags description started with "Exuberant Ctags ...", and the
"exuberant" part is not mentioned elsewhere. Kept it in a parenthesis
at the end of description.
* nix has the description "The Nix Deployment System". Since that
doesn't really say much what it is/does (especially after removing
the package name!), I changed that to "Powerful package manager that
makes package management reliable and reproducible" (borrowed from
nixos.org).
* Tons of "GNU Foo, Foo is a [the important bits]" descriptions
is changed to just [the important bits]. If the package name doesn't
contain GNU I don't think it's needed to say it in the description
either.
