[v13.x backport] fs: fix WriteStream autoClose order #32163
Conversation
ronag
force-pushed
the
backport-31790-to-v13.x
branch
from
6b59787
to
93889a5
Compare
ronag
force-pushed
the
backport-31790-to-v13.x
branch
from
93889a5
to
d696dbb
Compare
|
hey @ronag thanks for putting this together. I noticed that you modified the original meta data of the commit in this PR to include the backport PR URL and removed the original sign offs. For backport PRs it would be best if you could leave all the original commit meta data in place. Our tooling relies on that information for generating changelogs and it is a little more work for us to dig up the original meta data vs simply adding the Backport-PR-URL when ladning. FWIW don't worry about it for this PR (or the others you just opened) I'll fix that when landing, it is easy enough. Very much appreciate the initiative here though! |
ronag
force-pushed
the
backport-31790-to-v13.x
branch
from
d696dbb
to
8519946
Compare
| ); | ||
| file.write('should not work anymore', common.mustCall((err) => { | ||
| assert.strictEqual(err.code, 'ERR_STREAM_WRITE_AFTER_END'); | ||
| })); |
There was a problem hiding this comment.
This needs to be updated since the timing changes, which is the purpose of the original fix.
WriteStream autoClose was implemented by manually calling .destroy() instead of using autoDestroy and callback. This caused some invariants related to order of events to be broken. Fixes: nodejs#31776 PR-URL: nodejs#31790 Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Backport-PR-URL: nodejs#32163
ronag
force-pushed
the
backport-31790-to-v13.x
branch
from
8519946
to
c515e98
Compare
|
Looks like OSX failures are consistent (compare the CI & retry OSX runs in above comment) Here's a run against the current v13.x-staging branch as comparison |
|
@MylesBorins The failing tests seem totally unrelated though? Waiting for the referenced CI run to complete for comparison. |
|
So test.sequential/test-timers-blocking-callback seems to be the difference, but again I can't see any relation to this PR. I'll run just one more CI to be sure. |
|
@ronag dgram uses streams doesn't it? maybe try rebasing against staging and running tests again? |
This PR only affects |
|
Then it might be flakes that are now properly marked. Could you try rebasing and running CI one more time? |
Yep, already done, see CI link above. |
WriteStream autoClose was implemented by manually calling .destroy() instead of using autoDestroy and callback. This caused some invariants related to order of events to be broken. Fixes: #31776 Backport-PR-URL: #32163 PR-URL: #31790 Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com>
|
Fails are definitely unrelated the majority are already marked flaky but one isn't. Opened a PR to mark flaky Landed in 7eed9d6 |
PR-URL: #31790
Checklist
make -j4 test(UNIX), orvcbuild test(Windows) passes