Re: [Pan-users] Amusing gtk3 bug in latest git :)

[Duncan]

walt posted on Tue, 02 Oct 2012 17:16:48 -0700 as excerpted:

> Hi Heinrich.  I just pulled from git [87dfb36c8] and found a bug (IMO)
> when --with-gtk3 is set (apparently required by --with-webkit).
> 
> Each time I started pan it immediately went 'offline' for no apparent
> reason before it would begin any task, like pulling new headers, e.g.
> 
> When I rebuilt pan without --with-gtk3 (and therefor without
> --with-webkit)
> I quickly saw the cause of my 'offline' bug:  a dialog box popped up,
> telling me that I have reached my 'download limit' and 'go offline' was
> checked by default :p
> 
> When I unchecked 'go offline' then pan behaved normally again.  I'm
> about to re-enable gtk3 and webkit to see what happens next.
> 
> I'll report back later, and thanks for your hard work, as always!

I saw that on gtk2 as well, but having read git whatchanged, I knew about 
the new (and still testing) download meter option, that's intended to 
track amount of data downloaded and allow one to regulate it based on 
what their provider's plan allows.

But from my testing (and now apparently yours as well), there's a rather 
big bug somewhere, because the default cap is 1024 MB aka 1 GB, and with 
only gmane active for only a few text groups, I first watched the meter 
(which I was curious about and tested right away) accumulate a few bytes 
and then more or less forgot about it...

UNTIL A BIT LATER, when I hit the cap trying to refresh headers!

There's no way I'd have done a full GB in that time, not on text groups 
on gmane, just updating a dozen groups or so, definitely less than 50 
text posts and I think still single-digits.

And if for some reason pan HAD gone crazy and downloaded a whole gig of 
data, I'd have almost certainly been on gmane's ban-list by now, and I'm 
not.

So there's a SERIOUS bug in that bytes downloaded calculation, or at 
least was, when it was first introduced.

After unchecking the warn and disconnect, I tried again, but it didn't 
"take" until I hit the reset button as well.  But the commit log DID say 
something to the effect of "experimental", and I figured it was just a 
"live-git-version" hiccup.  Those do happen sometimes...   I had actually 
forgotten about it again until I saw this post.


Anyway, if the warning dialog popped up, it didn't "raise".  I didn't see 
it until I went investigating why my update wasn't happening, checked the 
task-list, and saw the offline, that kept going right back offline when I 
told it to online.  Then I remembered the meter mention in the log, 
decided to click it in the statusbar, and *THEN* got the dialog.

So there's still at least a couple bugs.  One is that the "warning" 
either doesn't popup, or doesn't raise if it does, so it's hidden behind 
the main pan window.

A second is some apparently VERY wrong math, somewhere.  Maybe it was 
1024 KB, not the 1024 MB that that the dialog stated as the cap.  That 
would have done it...


So I'm guessing you're not seeing a gtk3 bug, but probably the no-popup 
bug I saw on gtk2 as well.  Very likely what's happening is that the 
warning isn't triggering immediately; it's just going offline.  Then when 
pan restarts, which it obviously did when you switched back to gtk2, it 
sees the over-cap and THEN pops up the warning -- on restart not when it 
actually happens and goes offline.  So you got the dialog/warning not 
because you switched back to gtk2, but because you restarted pan, and the 
gtk2 switch just happened to occur at the same time. =:^\


Meanwhile, you're running the git version without even tracking the 
changes between updates?  Where's the fun in that?  Following git 
whatchanged is half the fun!  =:^)

(Tho I'll admit I don't read the whole thousands of entries commit log 
for the kernel merge window, I rely on community news coverage, make 
oldconfig, and google, for coverage there.)

The typically half-dozen commits between my pan updates, ABSOLUTELY I 
read git what-changed, and often git show the sources for individual 
commits that look interesting, too.  This thread is a good case in point 
as to why!  When pan started refusing to stay online, I immediately had a 
good suspect as to why, and knew exactly what new feature to go looking 
for the disable switch on! =:^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman