|
Ben Gamari
@
bgamari
New Hampshire, USA, Earth
|
|
I work on the Glorious Glasgow Haskell Compiler at @WellTyped.
|
|
|
614
Tweetovi
|
162
Pratim
|
1.298
Osobe koje vas prate
|
| Tweetovi |
|
Ben Gamari
@bgamari
|
8 h |
|
That's the one!
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
At this point it is looking likely that GHC will be moving to LLVM as its native toolchain on Windows; this will hopefully allow us to bypass many of the path and process issues that we have been historically fought with.
Current plan can be found here: gitlab.haskell.org/ghc/ghc/issues…
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
Perhaps the most shocking discovery of this entire ordeal is that it is literally *not possible* to reliably invoke the mingw toolchain we've been using for the last decade as it uses to its use of `exec`.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
Thankfully, I have the benefit of Phyx, GHC's resident Windows expert; however, even with his help trying to get GHC's Windows CI infrastructure to reliably pass has taken weeks and the end still seems a point on the horizon.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
Between vague documentation, decades of compatibility hacks, hacks on top of those hacks, a fundamentally different file and process model to POSIX, and the occasional toolchain bug, the amount of effort necessary to truly *support* Windows seems nothing short of herculean.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
I'm not sure how many people appreciate just how much of a maintenance challenge Windows represents for many open-source projects. Before I started working on GHC I had assumed it was bad.
However, now I know; no, it's much much worse.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
Thanks Apple. /s
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
What does the cited website say? Let's see:
When refusing SMTP transactions, our mail servers generate SMTP errors with information .... To find the error message, check your mail logs. Most of the errors will also include a URL that provides more information.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
Well investigating a contributor's messages to ghc-devs@haskell.org bounce I found this helpful message from Apple:
[CS01] Message rejected due to local policy. Please visit support.apple.com/en-us/HT204137
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
I initially had worried that home-manager would be too heavy-weight but frankly I have been pleasantly surprised; the cognitive overhead is quite minimal and for the most part things Just Work.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
`perf annotate` can do this.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
Absolutely true. It's a shame that there is duplication between the two.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
Previously I had a cobbled-together a set of shell scripts to approximate this, but it was nowhere near as reliable or convenient as home-manager and consequently I eventually just stopped using it.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
With home-manager I can make myself at home on such a machine by simply `git clone git.smart-cactus.org/ben/ben-env $HOME/.env` and running `home-manager build`. Log-out and log back in and I can go on my merry way, forgetting I'm even on a foreign machine.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
A nice advantage of home-manager is that you can use it on non-NixOS machines. I find myself often using such machines, often without even having root access (but with the ability to request installation of nix).
|
||
|
|
||
|
Ben Gamari
@bgamari
|
1. velj |
|
I use it and enjoy it. I just wish that the systemd modules were as complete as those in nixpkgs; otherwise I find it to be a great improvement over stuffing user configuration into /etc/nixos/configuration.nix.
|
||
|
|
||
| Ben Gamari proslijedio/la je tweet | ||
|
Joachim Breitner
@nomeata
|
30. sij |
|
The GHC Steering Committee just accepted the proposal “Unlifted Datatypes” by @sgraf1337:
github.com/ghc-proposals/…
|
||
|
|
||
|
Ben Gamari
@bgamari
|
28. sij |
|
I suspect this would be fixed by gitlab.haskell.org/ghc/ghc/merge_… but time is in short supply to finish this.
Anyways, I thought the ticket was quite timely.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
28. sij |
|
Heh. After sending that last tweet I started my morning ticket triage session and was greeted by gitlab.haskell.org/ghc/ghc/issues…. It seems that we aren't in the clear on ARM quite yet afterall.
|
||
|
|
||
|
Ben Gamari
@bgamari
|
28. sij |
|
Indeed, we would certainly welcome that conversation.
|
||
|
|
||