crash picture going into 3.0b1

0 views
Skip to first unread message

Wayne Mery

unread,
Nov 25, 2008, 12:49:22 AM11/25/08
to
Our users will be in a better place if we can fix the worst crashers
before shipping 3.0b1. We are behind in addressing top crashers for
3.0b1 because prior to this week it was impossible to rank trunk crashes
- due to lack of build symbols and crash server issues**.

For reasons stated below**, we must use custom reports to get a combined
picture of 3.0a3 and trunk:
-
http://crash-stats.mozilla.com/?do_query=1&product=Thunderbird&version=Thunderbird%3A3.0a3&version=Thunderbird%3A3.0b1pre&query_search=stack&query_type=contains&query=&date=&range_value=5&range_unit=days

omitting eudora crashers. and sqlite3_sql, which is Gloda and therefore
probably not applicable to b1 yields ...

Combined Ranking
1. imgRequest::NotifyProxyListener - Bug 457496
2. @0x14c0 (#1 on trunk)
3. nsTreeSelection::GetSingle - bug 456871
4. nsCOMPtr_base::assign_with_AddRef - Bug 445372
5. nsINode::GetCurrentDoc
6. nsImapMailFolder::NormalEndHeaderParseStream
7. nsProxyObjectCallInfo::RefCountInInterfacePointers - Bug 466318
(#2 on trunk)
8. nsCOMPtr_base::~nsCOMPtr_base - Bug 445372?
9. nsMailDatabase::UpdateFolderFlag - bug 450479 FIXED

Action list
1. 457496 hopefully will land Monday, with dmose or someone pushing.
2. @0x14c0 might mostly be
nsProxyObjectCallInfo::RefCountInInterfacePointers. Bug 466318 needs
work to get fixed by Monday.
3. bug 456871 - needs a dev?
4. Bug 445372 - needs analysis
5. nsINode::GetCurrentDoc - needs a bug
6. nsImapMailFolder::NormalEndHeaderParseStream - needs a bug
7. (see #2)
8. unclear if this is really Bug 445372, or something newer

In future few days the figures on crash stats should become more
representative, and getting Mac symbols and work on the crash server may
also help the rankings may change**. But I think the above is a good
hit list for triagers and developers to start focusing on - perhaps
nsProxyObjectCallInfo::RefCountInInterfacePointers in particular.

** Some problems remain but the situation is getting much better due to
work by gozer and many others. Windows symbols returned with
2008-11-20's build (ref bug 464247) after being gone for several weeks,
so we now have almost 6 days of good crash information. Mac symbols are
still lacking due to bug 445090/bug 466385. Canned topcrash queries
from the "top crash" button are currently not "accurate" due to Bug
461977 Duplicate Signatures in top crashers db; number crashes not
aggregated. Not also the Penelope group, whose crashers obviously
appear in the queries, are also working to get symbols so that those
crashes can be properly ranked and identified.

Wayne Mery

unread,
Nov 25, 2008, 12:13:50 PM11/25/08
to
On 11/25/2008 12:49 AM, Wayne Mery wrote:
> Our users will be in a better place if we can fix the worst crashers
> before shipping 3.0b1. We are behind in addressing top crashers for
> 3.0b1 because prior to this week it was impossible to rank trunk crashes
> - due to lack of build symbols and crash server issues**.
>
> For reasons stated below**, we must use custom reports to get a combined
> picture of 3.0a3 and trunk:
> -
> http://crash-stats.mozilla.com/?do_query=1&product=Thunderbird&version=Thunderbird%3A3.0a3&version=Thunderbird%3A3.0b1pre&query_search=stack&query_type=contains&query=&date=&range_value=5&range_unit=days
>
>
> omitting eudora crashers. and sqlite3_sql, which is Gloda and therefore
> probably not applicable to b1 yields ...
>
> Combined Ranking
> 1. imgRequest::NotifyProxyListener - Bug 457496
> 2. @0x14c0 (#1 on trunk)
> 3. nsTreeSelection::GetSingle - bug 456871

str also in bug 463560 Crash when selecting recipient from contacts sidebar

Wayne Mery

unread,
Nov 25, 2008, 12:18:08 PM11/25/08
to
On 11/25/2008 12:49 AM, Wayne Mery wrote:
> Our users will be in a better place if we can fix the worst crashers
> before shipping 3.0b1. We are behind in addressing top crashers for
> 3.0b1 because prior to this week it was impossible to rank trunk crashes
> - due to lack of build symbols and crash server issues**.
>
> For reasons stated below**, we must use custom reports to get a combined
> picture of 3.0a3 and trunk:
> -
> http://crash-stats.mozilla.com/?do_query=1&product=Thunderbird&version=Thunderbird%3A3.0a3&version=Thunderbird%3A3.0b1pre&query_search=stack&query_type=contains&query=&date=&range_value=5&range_unit=days
>
>
> omitting eudora crashers. and sqlite3_sql, which is Gloda and therefore
> probably not applicable to b1 yields ...
>
> Combined Ranking
> 1. imgRequest::NotifyProxyListener - Bug 457496
> 2. @0x14c0 (#1 on trunk)
> 3. nsTreeSelection::GetSingle - bug 456871
> 4. nsCOMPtr_base::assign_with_AddRef - Bug 445372
> 5. nsINode::GetCurrentDoc
>...

> 5. nsINode::GetCurrentDoc - needs a bug

only seen in 3.0a3 crashes, so perhaps fixed by Bug 447835 - Crash [@
nsINode::GetCurrentDoc] while trying to print/print preview with iframe
that has documentElement removed

Mark Banner

unread,
Nov 25, 2008, 2:11:39 PM11/25/08
to

I now have a fix posted for these (they are the same issue).

Standard8

Wayne Mery

unread,
Nov 28, 2008, 9:52:55 AM11/28/08
to
On 11/25/2008 12:49 AM, Wayne Mery wrote:

for #8, which is now the #1 topcrash, filed new Bug 467082 crash [@
nsCOMPtr_base::assign_with_AddRef(nsISupports*)] - enigmail

Reply all
Reply to author
Forward
0 new messages