1438455 - Memory leak in Thunderbird 59.0b1 leading to hang, on Debian 10 GNU/Linux

Status: RESOLVED DUPLICATE of bug 1439393

(MailNews Core :: Feed Reader, defect)

Description

[bodqhrohro]

User Agent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/64.0.3282.143 Safari/537.36 Vivaldi/1.95.1077.45

Steps to reproduce:

I upgraded 58.0b3 to 59.0b1, restarted TB and filled in passwords.


Actual results:

In several minutes after start it ate 1.5 GB of RAM, causing system hang, so I killed it and rolled back to 58.


Expected results:

TB 58 normally consumes 250–450 MB RAM so this looks like a terrible leak in the 59 version. How do I investigate what is it caused with?

Comment 1

[Wayne Mery (:wsmwk)]

> In several minutes after start it ate 1.5 GB of RAM, causing system hang, so I killed it and rolled back to 58.

Is this 59.0b1 supplied by debian?  Or from mozilla download?
Does it reproduce with Thunderbird started in safe mode?


Example reports of the last 5 years https://mzl.la/2o7zjeY - the most recent being bug 1322233, bug 1325825, bug 1410631
Many are unsolved in part to lack of participation from the reporter.

Comment 2

[Carsten Schoenert]

Debian hasn't a version 59.0b1 in any of the releases, the Debian version needs to be a bit adopted and would be 59.0~b1. Debian experimental has currently 58.0~b3-1, so I guess the reporter is talking about Mozilla releases running on a Debian testing. And I don't know I can upload 59.0~b1-1 in the next days due time constraints.

https://packages.debian.org/source/experimental/thunderbird

I haven't seen such issues while using binary packages from Mozilla.

Comment 3

[bodqhrohro]

(In reply to Wayne Mery (:wsmwk) from comment #1)
> Is this 59.0b1 supplied by debian?  Or from mozilla download?
58.0b3 was unpacked from tarball from thunderbird.net, and upgraded to 59.0b1 with TB's builtin updater.
> Does it reproduce with Thunderbird started in safe mode?
Yes.

Comment 4

[Wayne Mery (:wsmwk)]

Does memory also grow, and then hang, if you start using command line switch  -offline  ?
Hang with high CPU, or zero?

Comment 5

[bodqhrohro]

(In reply to Wayne Mery (:wsmwk) from comment #4)
> Does memory also grow, and then hang, if you start using command line switch
> -offline  ?
No, it stably consumes ~275 MB RES in offline mode.
> Hang with high CPU, or zero?
Hanging affects the whole system because the fast memory consumption (about 20–40 MB/s) causes thrashing and swapping when free memory runs out. TB itself performs well while there is enough free memory, however consuming CPU a lot too.

Comment 6

[marvinhk]

the same happens to thunderbird 59.0b2 on windows 7.

Comment 7

[marvinhk]

duplicate of bug 1353704 maybe?

crash details are listed here
https://bugzilla.mozilla.org/show_bug.cgi?id=1353704#c11

Comment 8

[Wayne Mery (:wsmwk)]

bodqhrohro,

Do you have RSS feeds?

Comment 9

[bodqhrohro]

(In reply to Wayne Mery (:wsmwk) from comment #8)
> bodqhrohro,
> 
> Do you have RSS feeds?

Yes, many of them.

Comment 10

[Wayne Mery (:wsmwk)]

If you disable the automatic update of all feeds, does the memory problem go away?

Comment 11

[bodqhrohro]

(In reply to Wayne Mery (:wsmwk) from comment #10)
> If you disable the automatic update of all feeds, does the memory problem go
> away?

Yes. And just when I enable it back, the memory consumption starts extremely growing again. The same happens on 59.0b2.

Comment 12

[Wayne Mery (:wsmwk)]

Same problem in version 60? (I would expect yes)

In bug 1439393  we re looking at potential RSS memory issue.

Comment 13

[bodqhrohro]

(In reply to Wayne Mery (:wsmwk) from comment #12)
> Same problem in version 60?

Yes.

If there are still only two reporters, maybe this happens with specific feeds only? I would try to make a binary search on a clean profile to find out a buggy one.

Comment 14

[Wayne Mery (:wsmwk)]

(In reply to bodqhrohro from comment #13)
> (In reply to Wayne Mery (:wsmwk) from comment #12)
> > Same problem in version 60?
> 
> Yes.
> 
> If there are still only two reporters, maybe this happens with specific
> feeds only? I would try to make a binary search on a clean profile to find
> out a buggy one.

For starters tTry
- https://archive.mozilla.org/pub/thunderbird/nightly/2017/11/2017-11-17-03-02-02-comm-central/thunderbird-59.0a1.en-US.win32.installer.exe
- https://archive.mozilla.org/pub/thunderbird/nightly/2017/11/2017-11-18-03-02-02-comm-central/thunderbird-59.0a1.en-US.win32.installer.exe

Comment 15

[Wayne Mery (:wsmwk)]

bodqhrohro 

nix comment 14. Try https://queue.taskcluster.net/v1/task/WQ8VMhVLT2uvvZOwuYnCsw/runs/0/artifacts/public/build/install/sea/target.installer.exe  Do you have good results?

Comment 16

[bodqhrohro]

I have no luck running all of three builds under Wine, they install successfully, but at start or nearly at start they crash. I will try on a VM later.

Comment 17

[marvinhk]

looks working fine with Thunderbird 60.0b2, up and running for 2 hours with low CPU and steady memory usage :)
http://ftp.mozilla.org/pub/thunderbird/releases/60.0b2/

Comment 18

[bodqhrohro]

(In reply to marvinhk from comment #17)
> looks working fine with Thunderbird 60.0b2, up and running for 2 hours with
> low CPU and steady memory usage :)
> http://ftp.mozilla.org/pub/thunderbird/releases/60.0b2/

For me too; on 60.0b2 the memory consumption stabilized at about 350 MB.

Comment 19

[alta88]

fixed in bug 1439393.  note that high cpu and memory if you've chosen to subscribe to a gigantic feed file is not a bug. there is only a bug if, when such processing completes, memory is not freed.