Uh oh!
There was an error while loading. Please reload this page.
[v10.x] Revert "http: always emit close on req and res" - #21809
Conversation
mcollina
commented
Jul 14, 2018
I would prefer if this is a @nodejs/tsc decision. Should we put it in the agenda for next week? |
targos
commented
Jul 14, 2018
Sure. Added the label. |
addaleax
commented
Jul 14, 2018
Node.js 10 has had the change for almost twice as long as it didn’t have it, and we’ve done security releases since then (i.e. we expect people to use versions that have it)… not knowing much about the subject matter, I’m not sure reverting is ideal?i |
mareksrom
commented
Jul 15, 2018
From different point of view - since 10.2.0 it was an undocumented API change, it became part of the documentation in version 10.6.0 just 10 days ago... How about projects relying on the previous documented version added in 0.6.7 released in January 2012? I think the number of them will be slighly bigger;) |
targos
commented
Jul 18, 2018
Note from TSC meeting: we also need to revert the documentation change. |
mhdawson
commented
Jul 18, 2018
From the discussion in the TSC meeting I'm +1 to reverting. |
mcollina
commented
Jul 18, 2018
This is the documentation change #21047. |
Trott
commented
Jul 18, 2018
Was there a resolution in the TSC meeting? Should the |
mcollina
commented
Jul 19, 2018
@Trott we did not have quorum. The proposed path is to revert this in Node 10, but leave it in in master. This approach would need necessarily 2 TSC signoffs because it's a retroactive semver-major change that should not have landed in master without 2 TSC signoffs. |
targos
commented
Jul 19, 2018
Doc change reverted |
targos
commented
Jul 19, 2018
CI: https://ci.nodejs.org/job/node-test-pull-request/15944/ Is this going to break |
mcollina
commented
Jul 19, 2018
@targos I do not think so. What broke eos was 8029a24. cc @mafintosh to confirm before landing. |
mafintosh
commented
Jul 19, 2018
|
targos
commented
Jul 20, 2018
CI is green. This needs at least one more @nodejs/tsc signoff. |
mcollina
commented
Jul 30, 2018
@nodejs/release @nodejs/lts I think this is ready to land. |
targos
commented
Aug 1, 2018
I'm going to add this to the release proposal at #22040 |
targos
commented
Aug 1, 2018
It's probably worth a sentence in the notable changes. Could someone help me write it? |
ruiaraujo
commented
Aug 1, 2018
Since this is already released, reverting it would just make a bad situation worse. SEMVER was already broken once with the initial release, reverting it would just break it again. |
targos
commented
Aug 1, 2018
I've added this back to the TSC agenda to discuss in today's meeting. |
MylesBorins
left a comment
There was a problem hiding this comment.
+1 reverting is the right move. This is something we should fix before LTS
There was a problem hiding this comment.
I feel the conflict here is which of the following we want to make migration from easier:
- Earlier v10.x versions
- Versions older than v10.x
Since v10.x is an LTS version, but we haven't reached LTS stage, there's still a lot of life remaining in v10.x as folks try to update to v10.x from versions older than v10.x.
It's a sucky situation, but I'm on the side of 👍 on this PR as well.
This reverts a commit that accidentally introduced a semver-major change to Node 10 and broke userland code. A subsequent fix to that change and documentation change are reverted with it. Revert "http: fix res emit close before user finish" This reverts commit 2a9c833. Revert "http: always emit close on req and res" This reverts commit 8029a24. Revert "doc: fix HTTP req/res 'close' description" This reverts commit 8ab7ea6. PR-URL: #21809 Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: Myles Borins <myles.borins@gmail.com> Reviewed-By: Rod Vagg <rod@vagg.org> Reviewed-By: Michael Dawson <michael_dawson@ca.ibm.com> Reviewed-By: Anatoli Papirovski <apapirovski@mac.com> Reviewed-By: Colin Ihrig <cjihrig@gmail.com> Reviewed-By: Tiancheng "Timothy" Gu <timothygu99@gmail.com>
targos
commented
Aug 1, 2018
Squashed and landed in 8799f43 |
This also reverts 2a9c833 which modifies the added file and line.
Refs: #20611 (comment)
/cc @nodejs/tsc
The change landed in v10.2.0. I suppose there is a risk to break code that is now relying on it?
Checklist
make -j4 test(UNIX), orvcbuild test(Windows) passes