[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Iceweasel 17.0 problem and solution

Hi Mikko,

Mikko Lehtinen wrote:
> "For Firefox, you could try creating an integer called
> toolkit.storage.synchronous
> in about:config, and setting it to 0.

Ah, there's a setting for that in Firefox? Didn't know that. Thanks
for that pointer!

> IIRC most of Firefox's problems on Linux are due to repeated fsync()
> calls, which on Linux mean "Sync everything to the disk RIGHT NOW!"

Yeah. Many applications do that on purpose. Especially those with
databases in the back, like Firefox. (Firefox uses SQLite databases
for bookmarks, history, password, etc.)

And I wouldn't be surprised if it's actually not the Firefox code but
the used SQLite libraries which contain the fsync calls.

> The upshot of setting toolkit.storage.synchronous to 0 is that Firefox
> should be less of a wallowing pig, but if it crashes you *might* lose some
> history entries, or bookmarks added during the session that crashed."

You might even loose *all* of your history entries, bookmarks,
cookies, extensions states, form history, and as far as I can seen
even stored passwords (signons.sqlite) if the according databases are
corrupt because of that after a powerloss. And I expect that this is
more riskier than most people want:


BTW, more or less the same effect can be achieved by using the tool
"eatmydata" from the package of the same name, e.g. calling "eatmydata
firefox", which is what I proposed to people asking about this issue
so far. (And that program is called "eatmydata" on purpose.)

The difference between toolkit.storage.synchronous and eatmydata is
likely that toolkit.storage.synchronous just affects the SQLite
databases while eatmydata affects the whole program, i.e. also the
cache. But I don't know if the cache code contains a lot of fsyncs.

> Could this be fixed somehow?

I wouldn't consider this a bug as it's obviously done on purpose and
most people do care about their bookmarks and especially stored

> Is it already fixed on Wheezy?

The current backports package (10.0.11esr-1~bpo60+1) is based on the
Wheezy package (10.0.11esr-1), so if you experience this in
Squeeze-Backports, you will experience this in Wheezy, too, unless the
changelog of the backports package says something different.

> For the backports version, it might be enough to just let the user
> aware of the problem and the solution.

That looks more like a general "solution", i.e. one that would not let
users run into unexpected data loss. (I think the package should be
quite conservative when it comes to data loss.) Since this is not
backports-specific, I'm cc'ing the Mozilla Packaging Team's mailing
list and I think further discussion should be done there.

		Regards, Axel
 ,''`.  |  Axel Beckert <abe@debian.org>, http://people.debian.org/~abe/
: :' :  |  Debian Developer, ftp.ch.debian.org Admin
`. `'   |  1024D: F067 EA27 26B9 C3FC 1486  202E C09E 1D89 9593 0EDE
  `-    |  4096R: 2517 B724 C5F6 CA99 5329  6E61 2FF9 CD59 6126 16B5

Reply to: