| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
- use common utils.nix code for generic test derivation;
- move spell checking code into standalone Bash file.
|
|
|
|
|
|
|
| |
Current unstable is marked as broken.
Also see:
- https://github.com/NixOS/nixpkgs/pull/61593
|
|
|
|
|
|
|
|
| |
The =sort= command requires $LANG to be set in order to sort consistently acrros
environments and OSes.
hunspell needs it to properly interpret non-ascii characters from the provided
custom dictionaries.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I couldn't find out what was causing the website build to fail.
I looked back in the git history and since I remembered the code back in January
1st 2019 was working, I searched for the last commit from that period (before I
started working on it again recently).
This was the perfect use case for using git bisect.
After identifying a good commit back then (
55e947d7901c17dbb1365a5cfb9f40acd50e903b), I read the =git help bisect=
instructions, I only had to find the command to combine with =git bisect run= to
test each commit.
Initially I considered doing a simple or:
nix-build -A subtasks.docs || nix-build -A build
This would work, but I later figured that the =publishScript= derivation existed
back then, and I could instead use it.
The final bisect commands were:
git bisect start HEAD 55e947d7901c17dbb1365a5cfb9f40acd50e903b --
git bisect run nix-build -A publishScript
git bisect reset
After the second command and around 30 seconds and a lot of output, git said:
error: build of '/nix/store/299c1mwcwxxl74cvzr13lzlfqfdqxdlc-publish.sh.drv' failed
997bed0f60a8d09fa14544c0d4c489f771e2534c is the first bad commit
commit 997bed0f60a8d09fa14544c0d4c489f771e2534c
Author: EuAndreh <eu@euandre.org>
Date: Sun May 19 01:39:07 2019 -0300
Split default.nix tasks
:100644 100644 00572eb6a32da956d18b14e1c00323bba1b34a72 31feb5afc45ed2592b8f7a81b4ba7fe56fbe3fd1 M default.nix
bisect run success
After looking at how default.nix was before and after the bad commit, I realised
that the =src= attribute of the =subtasks.docs= derivation was adapted
incorrectly after being changed from the =build= derivation. After fixing it, it
worked :)
Great job git :)
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Also, remove custom =nix-shell= shebang from Perl files. Instead, we should
inherit Perl from the =shell= derivation in default.nix.
|
|
|
|
|
|
| |
Instead of generating the HTML manually inside Emacs (=, e e h h=) and than
commiting the file to the repo, we now use a more generic approach with pandoc,
not tied to any Elisp configuration.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|