Earlier today (2026-03-30), all unmerged Gerrit patches to `mediawiki/core` that were both attached to a Phabricator task & were 2+ years old appear to have been abandoned _en masse_: [0]
There was a small amount of discussion after-the-fact in IRC (#wikimedia-releng): [1]
I am emailing wikitech-l to announce this myself, as it does not seem like this was done by the folks involved in making/actioning this decision. Speaking personally, I am saddened that it is apparently the case that there was no prior communication about today's action to the Wikimedia technical community, and that there was no (or, at least, no _recent_) consultation with the wider technical community prior to this highly impactful mass-action taking place.
From the current Gerrit search results at [2] (as viewed while typing this email - I'm not sure if Gerrit's search allows for filtering by the time a patch was abandoned), it appears that this action may have affected ~500-530 individual patches.
There was a Phabricator task filed in May 2025 that proposed the automatic abandonment of old changes; however, at the time of this action being taken, there had been no comments on that task since May 2025. [3]
Speaking personally again -- I wish that this hadn't happened. I wish there had been (wider, and more recent) community consultation. I wish that - at the very least - the voices expressing concern in the 'auto-abandonment' task [3] had been listened and responded to. That they weren't saddens me greatly: why should people take the time to express their concerns regarding suggested wide-scale changes such as this, if such concerns won't even be acknowledged prior to this sort of mass-action being taken? Given (e.g.) the thesis quoted in Tyler Cipriani's comment at [4], I am also especially concerned about what this sort of action may mean for the future health of MediaWiki & other Wikimedia software projects.
I do now hope that these changes can be mass-restored, in the same way that they were mass-abandoned; at least, until further community discussion on the matter can take place. (But that, I recognise, is an opinion that not everyone may necessarily share.)
---
[0] https://sal.toolforge.org/log/QS3ePp0B8tZ8Ohr0YH3W [1] https://wm-bot.wmcloud.org/browser/index.php?start=03%2F30%2F2026&end=03...) [2] https://gerrit.wikimedia.org/r/q/project:mediawiki/core+is:abandoned+before:... [3] https://phabricator.wikimedia.org/T393376 [4] https://phabricator.wikimedia.org/T393376#10831729
---
-- a smart kitten https://phabricator.wikimedia.org/p/A_smart_kitten/
I think we do have a problem with patchsets that have bitrotted and no longer effectively apply, but at the same time what are you supposed to do if you believe your patch is still a good patch, nobody has ever commented on it, and there is nothing obviously wrong with it. Its honestly not that unusual in our community for it to take months to get a patch merged.
-- Brian
On Monday, 30 March 2026, a smart kitten via Wikitech-l < wikitech-l@lists.wikimedia.org> wrote:
Earlier today (2026-03-30), all unmerged Gerrit patches to `mediawiki/core` that were both attached to a Phabricator task & were 2+ years old appear to have been abandoned _en masse_: [0]
There was a small amount of discussion after-the-fact in IRC (#wikimedia-releng): [1]
I am emailing wikitech-l to announce this myself, as it does not seem like this was done by the folks involved in making/actioning this decision. Speaking personally, I am saddened that it is apparently the case that there was no prior communication about today's action to the Wikimedia technical community, and that there was no (or, at least, no _recent_) consultation with the wider technical community prior to this highly impactful mass-action taking place.
From the current Gerrit search results at [2] (as viewed while typing
this email - I'm not sure if Gerrit's search allows for filtering by the time a patch was abandoned), it appears that this action may have affected ~500-530 individual patches.
There was a Phabricator task filed in May 2025 that proposed the automatic abandonment of old changes; however, at the time of this action being taken, there had been no comments on that task since May 2025. [3]
Speaking personally again -- I wish that this hadn't happened. I wish there had been (wider, and more recent) community consultation. I wish that
- at the very least - the voices expressing concern in the
'auto-abandonment' task [3] had been listened and responded to. That they weren't saddens me greatly: why should people take the time to express their concerns regarding suggested wide-scale changes such as this, if such concerns won't even be acknowledged prior to this sort of mass-action being taken? Given (e.g.) the thesis quoted in Tyler Cipriani's comment at [4], I am also especially concerned about what this sort of action may mean for the future health of MediaWiki & other Wikimedia software projects.
I do now hope that these changes can be mass-restored, in the same way that they were mass-abandoned; at least, until further community discussion on the matter can take place. (But that, I recognise, is an opinion that not everyone may necessarily share.)
[0] https://sal.toolforge.org/log/QS3ePp0B8tZ8Ohr0YH3W [1] https://wm-bot.wmcloud.org/browser/index.php?start=03% 2F30%2F2026&end=03%2F30%2F2026&display=%23wikimedia- releng#:~:text=!log%20gerrit%3A%20abandon,Txxxx%60) [2] https://gerrit.wikimedia.org/r/q/project:mediawiki/core+is: abandoned+before:%222026-03-30+14:00:00%22+after:%222026- 03-30+12:50:00%22,500 [3] https://phabricator.wikimedia.org/T393376 [4] https://phabricator.wikimedia.org/T393376#10831729
-- a smart kitten https://phabricator.wikimedia.org/p/A_smart_kitten/ _______________________________________________ Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l. lists.wikimedia.org/
Would it be too crazy to hire one staff member with the specific task of reviewing and merging community submitted patches? Seems to me like it would foster the developer community, and maybe even get more stuff done than if that person were coding directly. What would be the right place to propose something like this?
El lun, 30 de mar de 2026, 7:17 p.m., Brian Wolff via Wikitech-l < wikitech-l@lists.wikimedia.org> escribió:
I think we do have a problem with patchsets that have bitrotted and no longer effectively apply, but at the same time what are you supposed to do if you believe your patch is still a good patch, nobody has ever commented on it, and there is nothing obviously wrong with it. Its honestly not that unusual in our community for it to take months to get a patch merged.
-- Brian
On Monday, 30 March 2026, a smart kitten via Wikitech-l < wikitech-l@lists.wikimedia.org> wrote:
Earlier today (2026-03-30), all unmerged Gerrit patches to `mediawiki/core` that were both attached to a Phabricator task & were 2+ years old appear to have been abandoned _en masse_: [0]
There was a small amount of discussion after-the-fact in IRC (#wikimedia-releng): [1]
I am emailing wikitech-l to announce this myself, as it does not seem like this was done by the folks involved in making/actioning this decision. Speaking personally, I am saddened that it is apparently the case that there was no prior communication about today's action to the Wikimedia technical community, and that there was no (or, at least, no _recent_) consultation with the wider technical community prior to this highly impactful mass-action taking place.
From the current Gerrit search results at [2] (as viewed while typing
this email - I'm not sure if Gerrit's search allows for filtering by the time a patch was abandoned), it appears that this action may have affected ~500-530 individual patches.
There was a Phabricator task filed in May 2025 that proposed the automatic abandonment of old changes; however, at the time of this action being taken, there had been no comments on that task since May 2025. [3]
Speaking personally again -- I wish that this hadn't happened. I wish there had been (wider, and more recent) community consultation. I wish that
- at the very least - the voices expressing concern in the
'auto-abandonment' task [3] had been listened and responded to. That they weren't saddens me greatly: why should people take the time to express their concerns regarding suggested wide-scale changes such as this, if such concerns won't even be acknowledged prior to this sort of mass-action being taken? Given (e.g.) the thesis quoted in Tyler Cipriani's comment at [4], I am also especially concerned about what this sort of action may mean for the future health of MediaWiki & other Wikimedia software projects.
I do now hope that these changes can be mass-restored, in the same way that they were mass-abandoned; at least, until further community discussion on the matter can take place. (But that, I recognise, is an opinion that not everyone may necessarily share.)
[0] https://sal.toolforge.org/log/QS3ePp0B8tZ8Ohr0YH3W [1] https://wm-bot.wmcloud.org/browser/index.php?start=03%2F30%2F2026&end=03... ) [2] https://gerrit.wikimedia.org/r/q/project:mediawiki/core+is:abandoned+before:... [3] https://phabricator.wikimedia.org/T393376 [4] https://phabricator.wikimedia.org/T393376#10831729
-- a smart kitten https://phabricator.wikimedia.org/p/A_smart_kitten/ _______________________________________________ Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org
https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
People have been proposing that for as long as i've been here. Usually where it falls down is that to be a good reviewer you usually have to be well versed in the specific area of the codebase. A new hire is generally not a good code reviewer out of the box, and almost nobody is a good at reviewing everything.
Although perhaps someone hired to be a shepard would work (sort of similar to what andre does for phab) Just soneone to help shepard things through the process and make sure things do not fall thrugh the cracks.
I think a lot of the problem is many people do not feel empowered to say no, so we end up wasting a lot of time on silence or vauge platitudes.
That said i do think the mw core team has promised to spend part of their time reviewing code to good results.
-- Brian
On Monday, 30 March 2026, Felipe Schenone schenonef@gmail.com wrote:
Would it be too crazy to hire one staff member with the specific task of reviewing and merging community submitted patches? Seems to me like it would foster the developer community, and maybe even get more stuff done than if that person were coding directly. What would be the right place to propose something like this?
El lun, 30 de mar de 2026, 7:17 p.m., Brian Wolff via Wikitech-l < wikitech-l@lists.wikimedia.org> escribió:
I think we do have a problem with patchsets that have bitrotted and no longer effectively apply, but at the same time what are you supposed to do if you believe your patch is still a good patch, nobody has ever commented on it, and there is nothing obviously wrong with it. Its honestly not that unusual in our community for it to take months to get a patch merged.
-- Brian
On Monday, 30 March 2026, a smart kitten via Wikitech-l < wikitech-l@lists.wikimedia.org> wrote:
Earlier today (2026-03-30), all unmerged Gerrit patches to `mediawiki/core` that were both attached to a Phabricator task & were 2+ years old appear to have been abandoned _en masse_: [0]
There was a small amount of discussion after-the-fact in IRC (#wikimedia-releng): [1]
I am emailing wikitech-l to announce this myself, as it does not seem like this was done by the folks involved in making/actioning this decision. Speaking personally, I am saddened that it is apparently the case that there was no prior communication about today's action to the Wikimedia technical community, and that there was no (or, at least, no _recent_) consultation with the wider technical community prior to this highly impactful mass-action taking place.
From the current Gerrit search results at [2] (as viewed while typing
this email - I'm not sure if Gerrit's search allows for filtering by the time a patch was abandoned), it appears that this action may have affected ~500-530 individual patches.
There was a Phabricator task filed in May 2025 that proposed the automatic abandonment of old changes; however, at the time of this action being taken, there had been no comments on that task since May 2025. [3]
Speaking personally again -- I wish that this hadn't happened. I wish there had been (wider, and more recent) community consultation. I wish that
- at the very least - the voices expressing concern in the
'auto-abandonment' task [3] had been listened and responded to. That they weren't saddens me greatly: why should people take the time to express their concerns regarding suggested wide-scale changes such as this, if such concerns won't even be acknowledged prior to this sort of mass-action being taken? Given (e.g.) the thesis quoted in Tyler Cipriani's comment at [4], I am also especially concerned about what this sort of action may mean for the future health of MediaWiki & other Wikimedia software projects.
I do now hope that these changes can be mass-restored, in the same way that they were mass-abandoned; at least, until further community discussion on the matter can take place. (But that, I recognise, is an opinion that not everyone may necessarily share.)
[0] https://sal.toolforge.org/log/QS3ePp0B8tZ8Ohr0YH3W [1] https://wm-bot.wmcloud.org/browser/index.php?start=03% 2F30%2F2026&end=03%2F30%2F2026&display=%23wikimedia- releng#:~:text=!log%20gerrit%3A%20abandon,Txxxx%60) [2] https://gerrit.wikimedia.org/r/q/project:mediawiki/core+is: abandoned+before:%222026-03-30+14:00:00%22+after:%222026- 03-30+12:50:00%22,500 [3] https://phabricator.wikimedia.org/T393376 [4] https://phabricator.wikimedia.org/T393376#10831729
-- a smart kitten https://phabricator.wikimedia.org/p/A_smart_kitten/ _______________________________________________ Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l. lists.wikimedia.org/
Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l. lists.wikimedia.org/
I know this is sacrilege for some, but in 2026 AI can be a decent reviewer, even with opensource models. I wouldn't give it a +2, but it could certainly help improve patches and reducing the load of human reviewers.
Best, Strainu
În mar., 31 mar. 2026 la 01:42, Brian Wolff via Wikitech-l < wikitech-l@lists.wikimedia.org> a scris:
People have been proposing that for as long as i've been here. Usually where it falls down is that to be a good reviewer you usually have to be well versed in the specific area of the codebase. A new hire is generally not a good code reviewer out of the box, and almost nobody is a good at reviewing everything.
Although perhaps someone hired to be a shepard would work (sort of similar to what andre does for phab) Just soneone to help shepard things through the process and make sure things do not fall thrugh the cracks.
I think a lot of the problem is many people do not feel empowered to say no, so we end up wasting a lot of time on silence or vauge platitudes.
That said i do think the mw core team has promised to spend part of their time reviewing code to good results.
-- Brian
On Monday, 30 March 2026, Felipe Schenone schenonef@gmail.com wrote:
Would it be too crazy to hire one staff member with the specific task of reviewing and merging community submitted patches? Seems to me like it would foster the developer community, and maybe even get more stuff done than if that person were coding directly. What would be the right place to propose something like this?
El lun, 30 de mar de 2026, 7:17 p.m., Brian Wolff via Wikitech-l < wikitech-l@lists.wikimedia.org> escribió:
I think we do have a problem with patchsets that have bitrotted and no longer effectively apply, but at the same time what are you supposed to do if you believe your patch is still a good patch, nobody has ever commented on it, and there is nothing obviously wrong with it. Its honestly not that unusual in our community for it to take months to get a patch merged.
-- Brian
On Monday, 30 March 2026, a smart kitten via Wikitech-l < wikitech-l@lists.wikimedia.org> wrote:
Earlier today (2026-03-30), all unmerged Gerrit patches to `mediawiki/core` that were both attached to a Phabricator task & were 2+ years old appear to have been abandoned _en masse_: [0]
There was a small amount of discussion after-the-fact in IRC (#wikimedia-releng): [1]
I am emailing wikitech-l to announce this myself, as it does not seem like this was done by the folks involved in making/actioning this decision. Speaking personally, I am saddened that it is apparently the case that there was no prior communication about today's action to the Wikimedia technical community, and that there was no (or, at least, no _recent_) consultation with the wider technical community prior to this highly impactful mass-action taking place.
From the current Gerrit search results at [2] (as viewed while typing
this email - I'm not sure if Gerrit's search allows for filtering by the time a patch was abandoned), it appears that this action may have affected ~500-530 individual patches.
There was a Phabricator task filed in May 2025 that proposed the automatic abandonment of old changes; however, at the time of this action being taken, there had been no comments on that task since May 2025. [3]
Speaking personally again -- I wish that this hadn't happened. I wish there had been (wider, and more recent) community consultation. I wish that
- at the very least - the voices expressing concern in the
'auto-abandonment' task [3] had been listened and responded to. That they weren't saddens me greatly: why should people take the time to express their concerns regarding suggested wide-scale changes such as this, if such concerns won't even be acknowledged prior to this sort of mass-action being taken? Given (e.g.) the thesis quoted in Tyler Cipriani's comment at [4], I am also especially concerned about what this sort of action may mean for the future health of MediaWiki & other Wikimedia software projects.
I do now hope that these changes can be mass-restored, in the same way that they were mass-abandoned; at least, until further community discussion on the matter can take place. (But that, I recognise, is an opinion that not everyone may necessarily share.)
[0] https://sal.toolforge.org/log/QS3ePp0B8tZ8Ohr0YH3W [1] https://wm-bot.wmcloud.org/browser/index.php?start=03%2F30%2F2026&end=03... ) [2] https://gerrit.wikimedia.org/r/q/project:mediawiki/core+is:abandoned+before:... [3] https://phabricator.wikimedia.org/T393376 [4] https://phabricator.wikimedia.org/T393376#10831729
-- a smart kitten https://phabricator.wikimedia.org/p/A_smart_kitten/ _______________________________________________ Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org
https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org
https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
I followed that example and cleaned the old patches for the Math extension. I looked at every change individually and hoped that I didn't touch a change that someone was still interested in. In that case restoring the change is just one click away. physikerwelt (Moritz Schubotz)
On Tue, 31 Mar 2026 at 10:25, Strainu via Wikitech-l < wikitech-l@lists.wikimedia.org> wrote:
I know this is sacrilege for some, but in 2026 AI can be a decent reviewer, even with opensource models. I wouldn't give it a +2, but it could certainly help improve patches and reducing the load of human reviewers.
Best, Strainu
Hi, for what it's worth, I started a talk page section related to this email thread on https://meta.wikimedia.org/wiki/Talk:Wikimedia_Foundation_Annual_Plan/2026-2... .
Pine🌲
On Mon, 2026-03-30 at 19:30 -0300, Felipe Schenone via Wikitech-l wrote:
Would it be too crazy to hire one staff member with the specific task of reviewing and merging community submitted patches?
Yes, because nobody can know any and every codebase to effectively test touched functionality or propose a better-suited approach, plus nobody knows every and any code stewards' / maintainers' agenda/plans whether a proposed patch is within project scope or not, or guess their risk / urgency / complexity assessment of some proposed change.
Plus no matter which proposals this thread may bring up, I am quite sure that they have been discussed before in one of the many previous threads about code review over the last decade++.
andre
(PS: Not sure why the sentence above explicitly says "community".)
I usually ask one of the code reviewers I personally know to review my core patches. I think everyone else can do this.
Regardless of the reasoning for and the worthiness of the mass abandonment,[0] the attitude of "we have an internal justification for this, I'll get around to writing a task about it eventually" displayed in #wikimedia-releng is absolutely, fundamentally, entirely incompatible with the Gerrit privilege policy,[1] the guidelines for a healthy code review culture,[2] the Universal Code of Conduct,[3] and the general principles that hold the Wikimedia movement together.
There was no emergency here. There was no need to rush. If this action is truly needed, then it is the responsibility for those advocating for it to stand behind their decision and convince others.
These actions reflect poorly on the Wikimedia Foundation, and I expect them to do better.
0. I need not repeat what others have already said. 1. https://www.mediawiki.org/wiki/Gerrit/Privilege_policy 2. https://www.mediawiki.org/wiki/Guidelines_for_a_healthy_code_review_culture 3. https://foundation.wikimedia.org/wiki/Policy:Universal_Code_of_Conduct 3.3(1) AntiCompositeNumber (they/them)
On Mon, Mar 30, 2026, at 19:12, Nariman Mammadli via Wikitech-l wrote:
I usually ask one of the code reviewers I personally know to review my core patches. I think everyone else can do this. _______________________________________________ Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
Hi All
I've restored the 533 abandoned changes.
Before changes were abandoned, we should have given folks a heads up. Sorry for that.
The mass abandon was to help keep Gerrit running smoothly[0]. In short, Gerrit spends a lot of cycles checking if a change can be merged everytime something merges. For busy projects with many old changes (like mediawiki/core), the problem is worse.
It's stressful to keep a busy git forge running smoothly. In this case, we let an acute problem override the need to clearly communicate surprising changes to our users. We'll do better.
Thanks and sorry – Tyler
[0]: https://phabricator.wikimedia.org/T393376
On Mon, Mar 30, 2026 at 1:50 PM a smart kitten via Wikitech-l wikitech-l@lists.wikimedia.org wrote:
Earlier today (2026-03-30), all unmerged Gerrit patches to `mediawiki/core` that were both attached to a Phabricator task & were 2+ years old appear to have been abandoned _en masse_: [0]
There was a small amount of discussion after-the-fact in IRC (#wikimedia-releng): [1]
I am emailing wikitech-l to announce this myself, as it does not seem like this was done by the folks involved in making/actioning this decision. Speaking personally, I am saddened that it is apparently the case that there was no prior communication about today's action to the Wikimedia technical community, and that there was no (or, at least, no _recent_) consultation with the wider technical community prior to this highly impactful mass-action taking place.
From the current Gerrit search results at [2] (as viewed while typing this email - I'm not sure if Gerrit's search allows for filtering by the time a patch was abandoned), it appears that this action may have affected ~500-530 individual patches.
There was a Phabricator task filed in May 2025 that proposed the automatic abandonment of old changes; however, at the time of this action being taken, there had been no comments on that task since May 2025. [3]
Speaking personally again -- I wish that this hadn't happened. I wish there had been (wider, and more recent) community consultation. I wish that - at the very least - the voices expressing concern in the 'auto-abandonment' task [3] had been listened and responded to. That they weren't saddens me greatly: why should people take the time to express their concerns regarding suggested wide-scale changes such as this, if such concerns won't even be acknowledged prior to this sort of mass-action being taken? Given (e.g.) the thesis quoted in Tyler Cipriani's comment at [4], I am also especially concerned about what this sort of action may mean for the future health of MediaWiki & other Wikimedia software projects.
I do now hope that these changes can be mass-restored, in the same way that they were mass-abandoned; at least, until further community discussion on the matter can take place. (But that, I recognise, is an opinion that not everyone may necessarily share.)
[0] https://sal.toolforge.org/log/QS3ePp0B8tZ8Ohr0YH3W [1] https://wm-bot.wmcloud.org/browser/index.php?start=03%2F30%2F2026&end=03...) [2] https://gerrit.wikimedia.org/r/q/project:mediawiki/core+is:abandoned+before:... [3] https://phabricator.wikimedia.org/T393376 [4] https://phabricator.wikimedia.org/T393376#10831729
-- a smart kitten https://phabricator.wikimedia.org/p/A_smart_kitten/ _______________________________________________ Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-leave@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
wikitech-l@lists.wikimedia.org