registerOutputs(): Don't call canonicalisePathMetaData() twice #4054
Conversation
|
So we had to When I first rewrote Also, out of curiosity, how did |
|
The EOF is because the daemon aborts with an assertion failure. |
|
Ah, I didn't see the |
|
Ah! I see, the now-deleted comment should've instead read "check if all files were owned by the build user", since it's making them no longer be owned by that user. Thanks! |
| @@ -10,13 +10,15 @@ outPath=$(nix-store -rvv "$drvPath") | |||
|
|
|||
| echo "output path is $outPath" | |||
|
|
|||
| (! [ -w $outPath ]) | |||
There was a problem hiding this comment.
I don't think one needs the (..) subshell for this?
There was a problem hiding this comment.
It is needed. From the bash manpage under set -e: "The shell does not exit ... if the command's return value is being inverted with !".
There was a problem hiding this comment.
Oh, thanks! In that cases I think I need to fix some tests of mine on branches :)
|
This pull request has been mentioned on NixOS Discourse. There might be relevant details there: |
Fixes #4021. It also fixes a bad format string failure when debug output is enabled.
@Ericson2314 The second call to
canonicalisePathMetaData()should be superfluous now since we're always callingcanonicalisePathMetaData()in the loop overdrv->outputsat the start ofregisterOutputs(). But maybe I'm missing something, please check.