1148638 - C-C TB: "Error: __noSuchMethod__ is deprecated" from |make mozmill| and |make xpcshell-tests|

Status: RESOLVED DUPLICATE of bug 1142337

(Thunderbird :: Instant Messaging, defect)

Description

[ISHIKAWA, Chiaki]

Starting this month I noticed appearances of an Error message
(JavaScript warning) during the local test of C-C TB (by running |make
mozmill| and |make xpcshell-tests| test suites) regarding

    Error: __noSuchMethod__ is deprecated

Here are two places which seems to use this feature:

JavaScript warning: file:///REF-OBJ-DIR/objdir-tb3/dist/bin/components/irc.js, line 1845: Error: __noSuchMethod__ is deprecated

PROCESS | 677 | JavaScript warning: resource://testing-common/mailnews/nntpd.js, line 411: Error: __noSuchMethod__ is deprecated

The former appears just once in |make mozmill| log, and the latter
appears several times in |make xpcshell-tests| log.

I googled and found the following post:

"PSA: __noSuchMethod__ is deprecated in SpiderMonkey"
https://groups.google.com/forum/#!topic/mozilla.dev.platform/0EkHgphxUo8

The warning/error comes from the change mentioned above.

In the follow-up to the message above, Joshua Cranmer says:

>FWIW, I've found that both the uses in comm-central are basically poor
>man's reimplementation of ES6 classes, so I don't think it's worth
>removing this code until ES6 classes land.

Maybe so.

I am filing this so that this won't be forgotten.

I tried to see if __noSuchMethod__ is used for real in
nntpd.js and I found it is.
If I comment out the use of __noSuchMethod__ in nntpd.js, I get NEW errors
from |make xpcshell-tests| a la

TEST-UNEXPECTED-FAIL | mailnews/news/test/unit/test_nntpGroupPassword.js |  - "MODE READER,GROUP test.filter,AUTHINFO user group2,AUTHINFO pass pass2,GROUP test.filter" == "MODE READER,GROUP test.filter,AUTHINFO user group2,AUTHINFO pass pass2,GROUP test.filter,XOVER 1-8"
		     ...

So the clever feature is used in a complex way, and we need to rewrite
nntpd.js


It is possible that [1] is addressing the issue in irc.js, but I am
not sure.

[1] Bug 1142337 - Implement NormalizedMap without __noSuchMethod__

TIA


BTW, I wonder if ES6 features have been already in javascript
interpreter (including this ES6 class feature)?

I ask this because in the log of my local |make mozmill| test suite
run,
I see so many ES6-related warning, and they have overwhelmed other
serious issues when I scan the log.

To wit, the following is an excerpt from my log summarizer's output
against the log of |make mozmill|:
The number preceding the warning lines are the frequency of the
appearance of each warning. They are sorted in the descending order of
the frequency.

--- begin quote ---

 [ ...]

 ========================================
 JavaScript strict warning
 jquery.js and jquery-ui.js are ignored.
 (non-mozmill part)
 ========================================

     35 JavaScript warning: chrome://messenger/content/newmailaccount/accountProvisioner.js, line 810: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
     35 JavaScript warning: chrome://messenger/content/newmailaccount/accountProvisioner.js, line 754: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
     35 JavaScript warning: chrome://messenger/content/newmailaccount/accountProvisioner.js, line 735: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
     35 JavaScript warning: chrome://messenger/content/newmailaccount/accountProvisioner.js, line 574: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
     35 JavaScript warning: chrome://messenger/content/accountcreation/sanitizeDatatypes.js, line 154: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
     10 JavaScript warning: chrome://messenger/content/about-support/prefs.js, line 115: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
     10 JavaScript warning: chrome://messenger/content/about-support/init.js, line 89: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
     10 JavaScript warning: chrome://messenger/content/about-support/accounts.js, line 86: JavaScript 1.6's for-each-in loops are deprecated; consider using ES6 for-of instead
      7 JavaScript error: chrome://messenger/content/cloudfile/addAccountDialog.js, line 205: TypeError: this._accountType is null
      5 JavaScript error: resource:///modules/activity/alertHook.js, line 48: NS_ERROR_UNEXPECTED: Component returned failure code: 0x8000ffff (NS_ERROR_UNEXPECTED) [nsIMsgMailNewsUrl.server]
      5 JavaScript error: chrome://messenger/content/folderDisplay.js, line 1098: TypeError: this.view.displayedFolder is null

      [ ... ]
--- end quote ---

In the above, eight most frequent warnings/errors about use of ES6
for-of constructs. The number 35 suggests that they appear as part of
initialization of test modules (there are 35 test directories and
|make mozmill| runs against these directories.)

Only at the 9th position, we begin to see potentially serious issues.

      7 JavaScript error: chrome://messenger/content/cloudfile/addAccountDialog.js, line 205: TypeError: this._accountType is null
      5 JavaScript error: resource:///modules/activity/alertHook.js, line 48: NS_ERROR_UNEXPECTED: Component returned failure code: 0x8000ffff (NS_ERROR_UNEXPECTED) [nsIMsgMailNewsUrl.server]
      5 JavaScript error: chrome://messenger/content/folderDisplay.js, line 1098: TypeError: this.view.displayedFolder is null

      ...
      
I am wondering if the strange display errors I occasionally see are
related to these less frequent (in |make mozmill|)  error/warnings.
But skipping ES6-related warnings/errors have become very bothersome.

I hope the transition to ES6 features will end soon completely so that
we can focus on the real problems in the applications.

I wonder, by the way, if anyone reads logs from these test suites?

TIA

Comment 1

[Joshua Cranmer [:jcranmer]]

(In reply to ISHIKAWA, Chiaki from comment #0)
> JavaScript warning:
> file:///REF-OBJ-DIR/objdir-tb3/dist/bin/components/irc.js, line 1845: Error:
> __noSuchMethod__ is deprecated

As far as I know, the chat people are waiting until es6 classes (particularly super support) lands before this use is deprecated. It's a textbook es6 class example.

> PROCESS | 677 | JavaScript warning:
> resource://testing-common/mailnews/nntpd.js, line 411: Error:
> __noSuchMethod__ is deprecated

The latter is gone now, at least on c-c trunk.

> BTW, I wonder if ES6 features have been already in javascript
> interpreter (including this ES6 class feature)?

You can track these bugs with bug 694100. ES6 classes are already implemented, but I've been told that they shouldn't be used even in Mozilla code until they hit a release branch.

> In the above, eight most frequent warnings/errors about use of ES6
> for-of constructs. The number 35 suggests that they appear as part of
> initialization of test modules (there are 35 test directories and
> |make mozmill| runs against these directories.)

You mean bug 1118263? These sorts of bugs mostly need someone to step up and do the changes.

> I wonder, by the way, if anyone reads logs from these test suites?

Generally not, unless they fail.

Comment 2

[:aceman]

(In reply to Joshua Cranmer [:jcranmer] from comment #1)
> > In the above, eight most frequent warnings/errors about use of ES6
> > for-of constructs. The number 35 suggests that they appear as part of
> > initialization of test modules (there are 35 test directories and
> > |make mozmill| runs against these directories.)
> 
> You mean bug 1118263? These sorts of bugs mostly need someone to step up and
> do the changes.

There are some people who like to do these kind of bugs. I am sure Chiaki could also fix some of these.
As I agree they are usually not hard, they just need somebody to do the changes.

> > I wonder, by the way, if anyone reads logs from these test suites?
> Generally not, unless they fail.
I often do, but mostly on local runs.
It is good that Chiaki also does it and files these bugs for them.

Comment 3

[:aceman]

So now that the accountProvisioner and about-support has its own bug and the noSuchMethod is gone or known to the right people, what is left in this bug to do?

Comment 4

[Wayne Mery (:wsmwk)]

(get thee out of untriaged component)

Comment 5

[Patrick Cloke [:clokep]]

I think the nntpd part was fixed already and just the irc part is left? That's bug 1142337.