pipe runnable_binary wrapper output #1441
pipe runnable_binary wrapper output #1441matt-sm wants to merge 2 commits intobazel-contrib:mainfrom
Conversation
8dd0ab1 to
ab2f617
Compare
c8d32b8 to
f684b22
Compare
|
@cerisier this is another PR for your review. We've been running our windows builds with this fix for a while now and haven't seen any more runnable_binary failures |
|
Can you try with this or the equivalent as the |
|
The problem with this idea is that Fails on first line |
As a follow up to 1d8e362, we are still experiencing intermittent occurances of this issue whereby the wrapper script is empty on windows, and have been unable to reliably repro. This error message will at least give the user a clue of what the actual problem is.
7bdddf5 to
ce4752e
Compare
|
During more testing on windows with multiple Changing to |
|
All of this sound more and more like a sandboxing issue. Do you have the sandbox enabled ? are your actions local only ? |
|
On windows there is essentially no sandboxing. This is reproducible in a test I have that looks like where there are 40 different binaries I'm creating tests for. The tmp_$$$$ fixes this. Even with a unique filename (or $@ directly), the redirect (>) still produced intermittent empty files on Windows. |
Follow up to 1d8e362.
We are still experiencing intermittent occurances of this issue whereby the wrapper script is empty on windows. Was finally able to repro with a target that had a dep on a large number of runnable binaries. Inexplicably, when I changed the redirect to a pipe, the target never failed. With a redirect I would still see intermittent failures.
Closest msys2 related bug: msys2/MSYS2-packages#1703 (comment)