344769 - [meta] debug builds corresponding to releases must not be allowed to trigger assertions and spew lots of warnings

Status: RESOLVED WORKSFORME

(Firefox Build System :: General, defect)

Description

[Eyal Rozenberg]

It is very inconvenient, difficult even, for a developer to work on fixing a bug in one of the Mozilla apps, when your screen fills up with WARNING, DANGER WILL ROBINSON, etc. messages on stderr whenever you run your build - and what's worse, assertions get triggered, interrupting the normal interaction with the UI and sometime even preventing the possibility of decently testing your code (e.g. when it involves timed events).

Now, granted, this can't be avoided completely, but I would think that at least releases should not be made until at least there are no assertions triggered during normal use on the major platform, and the number of warning messages is reduced to a minimum - with this minimum being described somewhere (addendum to the release notes?)

I believe this would make contributing developers' lives a whole lot easier.

Comment 1

[Jesse Ruderman]

Can you paste some of the warnings you're encountering, or give bug numbers for them?

Comment 2

[Eyal Rozenberg]

Attachment: tbird warnings on startup

This is from a tbird build from the 1.5.0.4 source tarball.

Comment 3

[Eyal Rozenberg]

I also get ASSERTION: Have global without context?: 'mScriptGlobals[lang_ndx] == nsnull' file e:/moztools/src/mozilla/content/xul/document/src/nsXUL PrototypeDocument.cpp, line 846 when I click the tree root of some of my mail accounts (in this case, IMAP), and when I first started tbird after building it, I got more assertion failures.

Comment 4

[Eyal Rozenberg]

Immediately after 'ignore'ing the previous assertion, I get:

ASSERTION: Lying nsIInterfaceRequestor implementation!: '*aResult', file e:/moztools/src/mozilla/content/base/src/nsXMLHttpRequest.cpp, line 2031

Comment 5

[Eyal Rozenberg]

And a third assertion:

ASSERTION: nsWeakReference not thread-safe: '_mOwningThread.GetThread() == PR_GetCurrentThread()', file e:/moztools/src/mozilla/xpcom/build/nsWeakReference.cpp, line 146

Comment 6

[Jesse Ruderman]

You had me confused by using "release" to mean branch (as opposed to trunk).  It's usually used to mean "non-debug" (as opposed to debug).  Resummarizing.

Most Mozilla developers on Windows have XPCOM_DEBUG_BREAK set to "warn" so that assertions won't bring up dialogs.  (Mozilla developers on other platforms don't have to, because that or something similar is the default on Windows and Mac.)  Making that change to your environment should take care of a lot of the annoyance of assertions and the interference with timers.

That said, console noise annoys the heck out of me too.  I have a similar metabug, bug 341986.

Comment 7

[Ioana (away)]

I see no specific request associated to the qawanted keyword here. Is there anything you need from QA for this bug?

Comment 8

[BugBot [:suhaib / :marco/ :calixte]]

The meta keyword is there, the bug doesn't depend on other bugs and there is no activity for 12 months.
:marco, maybe it's time to close this bug?

Comment 9

[Marco Castelluccio [:marco]]

Probably no longer a problem.