74301 - New cache not properly updating object size

Status: RESOLVED FIXED

(Core :: Networking: Cache, defect)

Description

[tenthumbs]

With an 0331 nightly build, the new cache appears not to update the size
of an object on reload. I have an SSI HTML page which changes size
constantly. The cached file appears to be the largest size ever seen so
Mozilla display garbage eventually.

Attached is a copy of the cached object. Note the multiple </HTML> tags.
Everything after the first one is junk but mozilla displays the junk.

Comment 1

[Boris Zbarsky [:bzbarsky]]

[e-mail], did you forget to attach the cached copy?

Comment 2

[tenthumbs]

Attachment: cached version of HTML SSI document

Comment 3

[tenthumbs]

I'm guessing mozilla is not truncating the disk file properly.

Comment 4

[Patrick C. Beard]

Please verify this with today's landing.

Comment 5

[tenthumbs]

It's still there with an 0407 nightly trunk build.

One interesting thing is that about:cache reports bogus size values for
objects that chage a lot. In my case, the largest test page was 1571
bytes but, after about 25 reloads, about:cache said the page was 17258
bytes.

Comment 6

[gordon]

Hmm...we can't get there from here.  In nsCacheEntryDescriptor.cpp, the 
nsOutputStreamWrapper::OnWrite(PRUint32 count) method has no inkling of the 
offset being written to, so it keeps adding the count to the end.

Obviously a major malfunction...

Comment 7

[gordon]

Attachment: set data size according to offset in nsTransportWrapper::OpenOutputStream

Comment 8

[gordon]

Fix checked in.  There is still a bug in that nsFileTransports won't truncate 
files properly, but we'll handle that as a separate bug.

Comment 9

[Boris Zbarsky [:bzbarsky]]

Is there a bug# for the nsFileTransports bug?

Comment 10

[Boris Zbarsky [:bzbarsky]]

*** Bug 86469 has been marked as a duplicate of this bug. ***

Comment 11

[gordon]

I just created bug 86474 to track the lack of truncation with nsFileTransports,
but the problem may be fixed for the cache by other means.

Comment 12

[Boris Zbarsky [:bzbarsky]]

Do these symptoms still appear?  That is, is getting bug 86474 fixed necessary
to prevent garbage appearing?

Comment 13

[Bradley Baetz (:bbaetz)]

*** Bug 91254 has been marked as a duplicate of this bug. ***

Comment 14

[tenthumbs]

Still there in an 07-18-21 Linux nightly but the behavior is somewhat different.
You now have to load a dynamic page, go off to some other pages, and then return
with the Back button.

Yes, it's still a truncate problem. The garbage seen is really there in the disk
cache.

Comment 15

[benc]

http://bonsai.mozilla.org/cvslog.cgi?file=mozilla/netwerk/cache/public/nsICacheVisitor.idl
rev 1.9

Comment 16

[benc]

sorry, wrong bug.

jetlaggged comment.

Comment 17

[gordon]

Ben, you're causing me extra work when you mess with dead bugs like this.