| 25 Apr 2026 |
samasaur | ah okay got some in TIME_WAIT. unsure why the reproducer didn't do anything but curling random sites did | 18:05:10 |
samasaur | but they still drop to 0. i guess it's fixed in 15.7.3 ? | 18:05:27 |
Ihar Hrachyshka | my machine was at 52d. I guess I haven't built anything in 3d. :) | 18:05:45 |
samasaur | what 26.x are you on specifically? | 18:06:12 |
samasaur | bc 15.7.3 is from back in december... | 18:06:32 |
Ihar Hrachyshka | $ sw_vers
ProductName: macOS
ProductVersion: 26.2
BuildVersion: 25C56
| 18:06:49 |
samasaur | same release day | 18:07:16 |
| 26 Apr 2026 |
Randy Eckenrode | emily, thanks for the reviews. I updated the source release bump to fix the identified issues and responded regarding the cp patch. | 04:00:44 |
Randy Eckenrode | Have you had a chance to take another look at https://github.com/NixOS/nixpkgs/pull/511070? If you think we should wait until after the freeze (for 26.11), that’s understandable. We can live with the workarounds for 26.05. | 04:03:04 |
Randy Eckenrode | Also, of course Apple would release the 26.4 source releases right when I’m ready to merge this update. | 04:03:20 |
Randy Eckenrode | And another ld64 update. | 04:04:02 |
Randy Eckenrode | It looks like it adds some VisionOS stuff. Meh. | 04:04:44 |
John Wiegley | emily: hello | 05:36:32 |
John Wiegley | Trying to dig deeper into that issue I reported earlier today. | 05:37:07 |
emily | it was on my list but it got very late, sorry about cutting it close. I'm going to get to it today | 11:00:26 |
Randy Eckenrode | There are some fixes in the 26.4 libiconv. I don’t know whether they address the Autoconf issue, but I’m going to try doing another source release update today after merging the 26.3 update. | 12:27:14 |
emily | look at what I found while digging into the libiconv stuff: https://gitweb.git.savannah.gnu.org/gitweb/?p=autoconf.git;a=commitdiff;h=e1bbc9b93cdff61d70719c224b37970e065008bb | 15:55:40 |
emily | we're pioneers | 15:55:48 |
Randy Eckenrode | I updated just Apple’s libiconv and disabled all checks. I’m currently boostrapping with a goal to see whether I can build man-db. | 16:02:34 |
Randy Eckenrode | Now if only every other build system did that (and GCC finally upstreamed __builtin_available support). | 16:03:44 |
Randy Eckenrode | * Now if only every other build system did that (and __builtin_available finally got upstreamed into GCC). | 16:04:32 |
emily | alright, left a comment on the iconv PR | 16:31:10 |
Randy Eckenrode | I’m inclined to drop it for now if we’re going to end up with just these handfuls of packages. It’s similar to the approach I took dropping the automatic -liconv that ended up being a handful of packages. | 16:35:27 |
Randy Eckenrode | It also appears that Apple is making an attempt at fixing some of the behaviors. man-db no longer hangs during its build (as of libiconv-114) even though its tests still fail (currently about half way there to see if 115.whatever fixes it). | 16:36:14 |
Randy Eckenrode | I do wish Apple would actually work a bit more with FreeBSD instead of just keeping their fixes separate in a mess of #ifdef __APPLE__. | 16:38:04 |
Randy Eckenrode | * I do wish Apple would actually work a bit more with FreeBSD instead of just keeping their fixes separate in a mess of #ifdef __APPLE__. They might even get some fixes to problems on their platform. | 16:38:18 |
emily | FWIW, autoreconfHook actually propagates gettext. it looks like the actual reason this ever fails for any package is https://gitlab.com/grobian/html2text/-/merge_requests/57:
We also need to make sure autoreconf recognizes that it needs to run autopoint to install the gettext macros, which it does heuristically by checking whether AM_GNU_GETTEXT_VERSION is used,
| 16:39:38 |
Randy Eckenrode | Or like the libffi stuff getting merged, so we don’t have to build a fork and can have the fixes upstream. | 16:39:38 |
spewdins | hey.
I need to have a serious conversation with somebody in Nix infra. This is driving me nuts.
all of the linux-only software, in nixpkgs, which SHOULD also be available on mac (assuming a wayland compositor exists), is always marked broken, and it is infuriating on a whole other level as someone who is making a wayland compositor for macOS.
all the GTK toolkit theming, QT theming, none of it is going to matter if I can't simply install the software on macOS.
I understand why it isn't compiled for darwin upstream. But, the amount of software which should work on macOS with patches, it drives me nuts. There are many libraries, which are organized as linux only, and they can simply be installed on macOS with a little more effort.
qtwayland for example, should work on macOS. we don't need this to be linux only. I don't know who to talk to about changing this. maybe a custom darwin branch or something. | 16:39:40 |
emily | which is so cursed. | 16:39:50 |