extract julia directly to tool path to maintain mtimes#196
extract julia directly to tool path to maintain mtimes#196IanButterworth merged 18 commits intomasterfrom
Conversation
672e081 to
1293a89
Compare
f869840 to
f8eb0f4
Compare
fae5794 to
a34a04a
Compare
This reverts commit a34a04a.
7ef79c7 to
b677f38
Compare
|
Can we add a test for this? E.g. do |
|
Added mtime test which fails on master #200 |
Co-authored-by: Dilum Aluthge <[email protected]>
e7870f5 to
b789ee5
Compare
b789ee5 to
dcc55a3
Compare
There was a problem hiding this comment.
Changes are reasonable to avoid having the tool-cache module from modifying the mtimes. My only concern is that the tc.cacheDir function is described as "Caches a directory and installs it to the tool cacheDir" so it is possible that a non-breaking change to tc.cacheDir such as making the tool path immutable would break this code. However, since the version of @actions/tool-cache is pinned until we update it we should be able to avoid such a scenario.
Co-authored-by: Curtis Vogt <[email protected]>
|
Can we remove the tests from here, and keep the tests in #200, and merge #200 separately? Reason: I am hoping that this PR will be a short-term solution; I'm hoping that the long-term solution will be for me to get a patch into upstream |
If we're changing |
|
Ok, #200 is updated. Probably best to merge that as failing first then rebase this |
|
This is great! Is this the last change needed to make |
|
Yeah, should be. |
|
I made a test release for 1.9.5 using with And all platforms avoid re-precompilation once cached |
Upstream |

A bit of a hack to get julia installed with original mtimes to help JuliaLang/julia#50667
The comments explain the reasoning.
Also should be a bit faster as it avoids the
cpaction.You can see here that 1.10.0 has the release mtimes
The added mtime test fails on master to illustrate the issue. #200
Closes #100
Closes #165