🧬 in associated task causes fatal error on patch demo
Closed, ResolvedPublicBUG REPORT
Actions

Assigned To

Authored By

	Silvan_WMDE
	Jul 30 2025, 11:00 AM

Description

Steps to replicate the issue (include links if applicable):

Go to https://patchdemo.wmcloud.org/

What happens?:
A fatal error is shown on the bottom of the page:

Fatal error: Uncaught mysqli_sql_exception: Incorrect string value: '\xF0\x9F\xA7\xAC G...' for column `patchdemo`.`tasks`.`title` at row 1 in /var/www/html/includes.php:376 Stack trace: #0 /var/www/html/includes.php(376): mysqli_stmt->execute() #1 /var/www/html/includes.php(214): get_task_data() #2 /var/www/html/index.php(232): get_wiki_data_from_row() #3 {main} thrown in /var/www/html/includes.php on line 376

I saw the same error on the 'create' page when trying to set up a demo for my patch.

What should have happened instead?:

no fatal error

Software version (on Special:Version page; skip for WMF-hosted wikis like Wikipedia):

Other information (browser name/version, screenshots, etc.):

Details

Related Changes in Gerrit:

	Subject	Repo	Branch	Lines +/-
	DNM: Patch for testing	mediawiki/core	master	+1 -0

Customize query in gerrit

Related Changes in GitLab:

	Title	Reference	Author	Source Branch	Dest Branch
	Use utf8mb4	repos/test-platform/catalyst/patchdemo!187	jhuneidi	utf8mb4	main

Customize query in GitLab

Related Objects

Mentioned Here: T193222: Error: 1071 Specified key was too long; max key length is 767 bytes
T400151: 🧬 GraphiQL

Event Timeline

Silvan_WMDE created this task.Jul 30 2025, 11:00 AM

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptJul 30 2025, 11:00 AM

FWIW, \xF0\x9F\xA7\xAC is the UTF-8 encoding of U+1F9EC DNA DOUBLE HELIX (🧬) if I’m not mistaken.

Wild guess: maybe the Patch Demo database is using utf8mb3 and needs utf8mb4 in order to accommodate a non-BMP emoji character?

ooh, that means I broke it when I tried to set up a patch demo for https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/1174403 (with our corresponding phab task T400151)

A_smart_kitten subscribed.Jul 30 2025, 1:05 PM

In T400801#11046571, @Lucas_Werkmeister_WMDE wrote:

FWIW, \xF0\x9F\xA7\xAC is the UTF-8 encoding of U+1F9EC DNA DOUBLE HELIX (🧬) if I’m not mistaken.

Wild guess: maybe the Patch Demo database is using utf8mb3 and needs utf8mb4 in order to accommodate a non-BMP emoji character?

Sure enough:

MariaDB [(none)]> show variables where Variable_name like 'character%' or Variable_name like 'collation%';
+--------------------------+--------------------------------------+
| Variable_name            | Value                                |
+--------------------------+--------------------------------------+
| character_set_client     | utf8mb3                              |
| character_set_collations |                                      |
| character_set_connection | utf8mb3                              |
| character_set_database   | utf8mb3                              |
| character_set_filesystem | binary                               |
| character_set_results    | utf8mb3                              |
| character_set_server     | utf8mb3                              |
| character_set_system     | utf8mb3                              |
| character_sets_dir       | /opt/bitnami/mariadb/share/charsets/ |
| collation_connection     | utf8mb3_general_ci                   |
| collation_database       | utf8mb3_general_ci                   |
| collation_server         | utf8mb3_general_ci                   |
+--------------------------+--------------------------------------+
12 rows in set (0.001 sec)

Fortunately, there's an entertaining post for us to follow: https://mathiasbynens.be/notes/mysql-utf8mb4

Note that the immediate issue seems to have been resolved (for me) by T400151#11046800, but this issue will still presumably need addressing for the future

ItamarWMDE subscribed.Jul 30 2025, 1:31 PM

Jakob_WMDE subscribed.Jul 30 2025, 1:34 PM

I didn't actually expect to resolve the issue by renaming the task, just did it in preparation of possibly re-creating the patch demo instance.

So this means, the fatal error occured on the attempt to write the 🧬 into the utf8mb3 database field. I'm a bit surprised it didn't only fail creating my wiki, but instead caused a general error on patch demo.

I'm glad this turned out to be a quick-and-easy fix.

utf8mb3 only allows characters up to 3 bytes which indeed excludes any characters. It means it only supports the Basic Multingual Plane and would rejects any characters 4 bytes utf-8 characters.

You can do some input filtering to reject inputs that is invalid utf-8 or 4 bytes characters. I am also not sure how the 4 bytes character managed to be inserted without MariaDB rejecting the INSERT statement, maybe the client is set to utf8mb4/binary?

For the context, MediaWiki / Wikimedia uses binary because at the time we used characters that were not implemented in MySQL. MediaWiki does the necessary filtering/validation and Debian defaults to utf8mb4 since Stretch (T193222), it allows Unicode characters of 4 bytes.

thcipriani renamed this task from Fatal error on patch demo to 🧬 in associated task causes fatal error on patch demo.Jul 30 2025, 9:00 PM

Change #1174553 had a related patch set uploaded (by Thcipriani; author: Thcipriani):

[mediawiki/core@master] DNM: Patch for testing

https://gerrit.wikimedia.org/r/1174553

gerritbot added a project: Patch-For-Review.Jul 30 2025, 9:00 PM

jhuneidi opened https://gitlab.wikimedia.org/repos/test-platform/catalyst/patchdemo/-/merge_requests/187

Use utf8mb4

In T400801#11046833, @A_smart_kitten wrote:

Note that the immediate issue seems to have been resolved (for me) by T400151#11046800, but this issue will still presumably need addressing for the future

Thanks for this! Lowering to high priority.

@jeena 's got a change that should do the heavy lifting. Getting late, so we'll deploy tomorrow. Planning to merge, run a final db backup, and run the script to alter tables on db.

thcipriani edited projects, added Catalyst (akesi npa); removed Catalyst (PatchDemo).Jul 31 2025, 4:07 PM

thcipriani moved this task from Backlog to In progress on the Catalyst (akesi npa) board.Jul 31 2025, 4:17 PM

kindrobot merged https://gitlab.wikimedia.org/repos/test-platform/catalyst/patchdemo/-/merge_requests/187

Use utf8mb4

We ran a database migration for patchdemo today.

We took notes on this etherpad.

For phab posterity, steps:

Final prod backup/final migration test on local dump of prod
Merge
Deploy to staging
Final dump of staging db
Create broken wiki on staging db
Run migration on staging db and confirm it's fixed
Steps 3–6, but on prod

All working, thanks for reporting!

Change #1174553 abandoned by Thcipriani:

[mediawiki/core@master] DNM: Patch for testing

Reason:

T400801 testing complete

https://gerrit.wikimedia.org/r/1174553

jnuche awarded a token.Aug 5 2025, 12:32 PM

Maintenance_bot removed a project: Patch-For-Review.Aug 5 2025, 1:31 PM

Change #1174553 restored by Thcipriani:

[mediawiki/core@master] DNM: Patch for testing

https://gerrit.wikimedia.org/r/1174553

gerritbot added a project: Patch-For-Review.Sep 25 2025, 4:11 PM

Test wiki created on Patch demo by TCipriani (WMF) using patch(es) linked to this task:
https://b8e9b7b175.catalyst.wmcloud.org/w/

Test wiki created on Patch demo by TCipriani (WMF) using patch(es) linked to this task:
https://43b2909e45.catalyst.wmcloud.org/w/

Test wiki on Patch demo by TCipriani (WMF) using patch(es) linked to this task was deleted:

~~https://b8e9b7b175.catalyst.wmcloud.org/w/~~

Test wiki on [[ | Patch demo ]] by TCipriani (WMF) using patch(es) linked to this task was deleted:

~~https://43b2909e45.catalyst.wmcloud.org/w/~~

Test wiki on Patch demo by TCipriani (WMF) using patch(es) linked to this task was deleted:

~~https://43b2909e45.catalyst.wmcloud.org/w/~~

🧬 in associated task causes fatal error on patch demoClosed, ResolvedPublicBUG REPORTActions

Description

Details

Related Objects

Event Timeline

🧬 in associated task causes fatal error on patch demo
Closed, ResolvedPublicBUG REPORT
Actions