Page MenuHomePhabricator

Add mathoid database table
Closed, ResolvedPublic

Description

Author: physik

Description:
Short description:
Add the mathoid table that stores MathML and SVG output for texvc markup.
It is seperated from the old math table called math.
Which wikis are affected:
all
Which tables:
mathoid
What is the change to those tables:
new table
Links to gerrit changes and/or other related bug reports:
https://gerrit.wikimedia.org/r/#/c/135521/
https://bugzilla.wikimedia.org/show_bug.cgi?id=65793


Version: unspecified
Severity: normal
See Also:
https://bugzilla.wikimedia.org/show_bug.cgi?id=65793
https://bugzilla.wikimedia.org/show_bug.cgi?id=66587

Details

Reference
bz66492

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 22 2014, 3:19 AM
bzimport set Reference to bz66492.
bzimport added a subscriber: Unknown Object (MLST).

physik wrote:

Report created via template from https://wikitech.wikimedia.org/wiki/Schema_changes
To be honest I don't understand why this table has to be created manually.

This bug is a Wikimedia-specific bug. There is already the Math extension specific bug at bug 65793.

This probably won't be done before next week so if there are backwards-incompatible changes committed in the Math extension, they should be reverted now, ie: before 10:00 UTC on Thursday the 12th.

Change 138993 had a related patch set uploaded by Physikerwelt:
Disable MathML rendering option

https://gerrit.wikimedia.org/r/138993

Change 139068 had a related patch set uploaded by Physikerwelt:
Disable MathML rendering option

https://gerrit.wikimedia.org/r/139068

Change 138993 abandoned by Physikerwelt:
Disable MathML rendering option

Reason:
moved Icf0c462b57af6e0fc0bc80994d07b004f2829f56

https://gerrit.wikimedia.org/r/138993

physik wrote:

OK. The change was not merged and no problem was reported.
Maybe there is no problem.
If there is a problem the change

https://gerrit.wikimedia.org/r/138993

will disable the feature that potentially causes a problem due to the missing mathoid table.

physik wrote:

OK. Now the new branch has been created and the expected can be produced.
"
Database error
A database query error has occurred. This may indicate a bug in the software.

Function: MathRenderer::readFromDatabase
Error: 1146 Table 'mediawikiwiki.mathoid' doesn't exist (10.64.16.8)
"
From
https://www.mediawiki.org/wiki/User:Physikerwelt
rendered with MathML mode.

Change 139068 merged by jenkins-bot:
Disable MathML rendering option

https://gerrit.wikimedia.org/r/139068

This has caused headaches all day. There is a reason we have a process for adding new tables in production. We don't just run install.php in production.

In the future, if you don't follow https://wikitech.wikimedia.org/wiki/Schema_changes we'll have to fully revert the math extension or completely turn it off. Continued recklessness in this area is not sustainable.

(In reply to Greg Grossmeier from comment #9)

This has caused headaches all day. There is a reason we have a process for
adding new tables in production. We don't just run install.php in production.

In the future, if you don't follow
https://wikitech.wikimedia.org/wiki/Schema_changes we'll have to fully
revert the math extension or completely turn it off. Continued recklessness
in this area is not sustainable.

(In reply to Ori Livneh from bug 65793 comment #15)

I reverted to 1bb3bfa3b5656 because the exception log was flooded with
errors related to the math extension. I also revoked 2 rights on
extension/Math from Physikerwelt and Frédéric Wang until Greg has had a
chance to review exactly what happened.

I may be turned around and confused about the order of events here and I'll be interested to read Greg's formal review.

That said, I get the impression that both of these quoted comments are placing the responsibility of deploying code and maintaining Wikimedia Foundation servers on the shoulders of two particular extension authors. This seems a bit... wrong and backward to me.

It looks to me like one of these two volunteers was kind enough to (promptly!) provide a Wikimedia-specific patch after being asked to, which he was under no obligation to do, and that patch then didn't get merged and deployed until some 16 hours later. This issue was causing problems "all day," eh?

If the Wikimedia Foundation isn't reviewing extension code before merging it into the release branches and shipping it to the servers, that sounds like a pretty major bug to me... but I'm not sure why people are calling the volunteer behavior here reckless and revoking access when it's the non-volunteer behavior here that seems to need investigation.

(In reply to MZMcBride from comment #10)

That said, I get the impression that both of these quoted comments are
placing the responsibility of deploying code and maintaining Wikimedia
Foundation servers on the shoulders of two particular extension authors.
This seems a bit... wrong and backward to me.

Anyone, volunteer or staff, with 2 access on MediaWiki core, WMF-deployed extensions, or other repositories that affect the WMF sites is expected to take part in ensuring that that code doesn't negatively affect the WMF sites. This is a requirement of having 2 access to those repositories, and is stated on https://www.mediawiki.org/wiki/Gerrit/+2

If someone does not want this responsibility, then they shouldn't have 2. We'll still welcome their contributions, just like the wiki communities welcome contributions from people who aren't admins or 'crats.

It would certainly be doable for only WMF staff to be allowed 2 to these repositories as you're saying should be the case, although it would slow down development in some areas. But why *shouldn't* we accept help in this from skilled volunteers as we do in so many other areas?

bug 66587 suggests "Leave it as TeX (for text browsers)" doesn't work either

Is there any reason the table can't be added now?

Change 139381 had a related patch set uploaded by Reedy:
Remove MW_MATH_SOURCE as errors without mathoid table

https://gerrit.wikimedia.org/r/139381

Change 139381 merged by jenkins-bot:
Remove MW_MATH_SOURCE as errors without mathoid table

https://gerrit.wikimedia.org/r/139381

physik wrote:

Can anyone comment on this comment I made on the other bug-ticket:
"
https://gerrit.wikimedia.org/r/#/c/138023/
I could image that the wmf8 code produces database write operation that are not required.
"
I think that could be relevant.

Change 139421 had a related patch set uploaded by Reedy:
Re-enable all Math modes

https://gerrit.wikimedia.org/r/139421

The table is still created, even though Gerrit Notification Bot reopened this :)

Change 135522 had a related patch set uploaded by Physikerwelt:
Add table for new math renderer Mathoid

https://gerrit.wikimedia.org/r/135522

Change 135522 merged by jenkins-bot:
Add table for new math renderer Mathoid

https://gerrit.wikimedia.org/r/135522

Change 153610 had a related patch set uploaded by Physikerwelt:
Re-enable all Math modes on beta

https://gerrit.wikimedia.org/r/153610

Change 139421 merged by jenkins-bot:
Re-enable all Math modes

https://gerrit.wikimedia.org/r/139421