Bug 19793 – Web Inspector should show compressed size for network resources

• Status: RESOLVED
• Resolution: FIXED
• Component: Web Inspector (Deprecated)

[Haris Pobric]

One thing I noticed is that the Web Inspector gives different results for the same page when used on Tiger and Leopard. Specifically, when analyzing resources, Web Inspector indicates the size of Prototype.js as ~31 KB when used on Tiger but ~123 KB when the same page is analyzed on a Leopard machine. The script is served via Google Ajax API. The behavior is persistent across different machines and web sites and servers.

I have noticed this behavior with the nightly builds for the past couple of weeks.

[Mark Rowe (bdash)]

I suspect that CFNetwork on Tiger is giving us the compressed size (eg, size over the wire) while Leopard is giving us the uncompressed size (after any gzip compression has been reversed).

[Mark Rowe (bdash)]

<rdar://problem/6041709>

[Alexey Proskuryakov]

I'm pretty sure we have a duplicate in Bugzilla, but I cannot find it now.

[Mark Janssen]

(In reply to comment #1)
> I suspect that CFNetwork on Tiger is giving us the compressed size (eg, size
> over the wire) while Leopard is giving us the uncompressed size (after any gzip
> compression has been reversed).

I came across this bug when enabling gzip on a web server didn't make any difference to the size shown in the Web Inspector on the latest Nightly on Leopard. I haven't got a copy of Tiger so I can't confirm the original bug report.

Maybe the best thing to do is to take the 'Content-Length' header and base reports on that.

[Robert Blaut]

(In reply to comment #3)
> I'm pretty sure we have a duplicate in Bugzilla, but I cannot find it now.
> 

Is it a bug 14279?

[Alexey Proskuryakov]

That's probably not what I was thinking of, but it is certainly related.

[Alexey Proskuryakov]

*** Bug 30795 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

*** Bug 38181 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

See also: bug 31019.

[Vsevolod Vlasov]

Created attachment 86054 [details]
Patch enabling inspector support for showing transfer size without Content-Length headers.

This patch enables inspector support for showing transfer size without Content-Length headers.
It is still responsibility of each port to provide valid transfer size data to ResourceLoader/ResourceHandleClient.

This bug should probably be renamed (if possible) to something more general, like "Web Inspector does not always show correct transfer size (for compressed/chunked data)", and then separate bugs could be created for each port.

[Vsevolod Vlasov]

Created attachment 86073 [details]
Patch follow-up

Some fixes added

[Pavel Feldman]

Comment on attachment 86073 [details]
Patch follow-up

View in context: https://bugs.webkit.org/attachment.cgi?id=86073&action=review

> LayoutTests/http/tests/inspector/network/network-size-chunked.html:16
> +    if (++loadedResourceCount === 1) {

I don't think you need a counter here.

> LayoutTests/http/tests/inspector/network/network-size-chunked.html:25
> +    InspectorTest.evaluateInPage("loadPages()");

Where is loadPages?

> LayoutTests/platform/chromium/test_expectations.txt:3190
> +BUG_VSEVIK SKIP : http/tests/inspector/network/network-size-chunked.html = FAIL

Could you please file a bug for all of these?

> Source/WebCore/inspector/front-end/Resource.js:244
> +        if (this._transferSize)

What about 0?

> Source/WebCore/loader/appcache/ApplicationCacheGroup.cpp:561
> +    InspectorInstrumentation::didReceiveContentLength(m_frame, m_currentResourceIdentifier, length, -1);

No traffic here.

> Source/WebCore/platform/network/mac/ResourceHandleMac.mm:926
> +    m_handle->client()->didReceiveData(m_handle, (const char*)[data bytes], [data length], -1);

Put a FIXME and a bug reference here.

[Vsevolod Vlasov]

Created attachment 86109 [details]
Patch with fixes

All done.

Added fixes for QT, GTK, Win.

[Alexey Proskuryakov]

Could you please write down what is being fixed in this patch? Will it fully work in Safari on Leopard and later?

[Vsevolod Vlasov]

Created attachment 86113 [details]
Patch with fixes2

I have uploaded my path again to resolve merge conflicts.

No, it will not make it fully work in any port.

This particular patch just makes it possible to pass actual transfer size to inspector.
Each port should be changed accordingly to pass this value in ResourceHandleClient. 
I am not sure but it could probably require some changes outside of WebKit.

I created separate bug for chromium port which I am currently working on: https://bugs.webkit.org/show_bug.cgi?id=56602
I suggest we create separate bugs for other ports.

[Vsevolod Vlasov]

Created attachment 86155 [details]
Patch + changelogs update forgotten earlier

[Pavel Feldman]

Comment on attachment 86155 [details]
Patch + changelogs update forgotten earlier

View in context: https://bugs.webkit.org/attachment.cgi?id=86155&action=review

Please send complete patch.

> LayoutTests/ChangeLog:5
> +        Added transfer size support in inspector for compressed data

Needs "Web Inspector:" prefix and the bug href.

> LayoutTests/ChangeLog:7
> +        Need a short description and bug URL (OOPS!)

OOPS breaks presubmit checks.

[Vsevolod Vlasov]

Created attachment 86157 [details]
Patch with fixes

[Vsevolod Vlasov]

Created attachment 86165 [details]
Patch

[Build Bot]

Attachment 86165 [details] did not build on win:
Build output: http://queues.webkit.org/results/8199301

[Alexey Proskuryakov]

> No, it will not make it fully work in any port.

Note that this bug is asking to make it work in Safari.

[Pavel Feldman]

Comment on attachment 86165 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=86165&action=review

> LayoutTests/ChangeLog:7
> +        Web Inspector: Web Inspector Gives Different Results on Tiger and Leopard (should show compressed size)

ChangeLog structure:

Reviewed By

Title
href

Description

* files

[Pavel Feldman]

(In reply to comment #21)
> > No, it will not make it fully work in any port.
> 
> Note that this bug is asking to make it work in Safari.

Yeah, we should move this patch to a different bug and make this one depend on it.

[Ilya Tikhonovsky]

Comment on attachment 86165 [details]
Patch

was landed as r81580

[Alexey Proskuryakov]

Ilya, why did you close this bug? Are you saying that Web Inspector now displays a correct size in Safari?

[Pavel Feldman]

Comment on attachment 86165 [details]
Patch

I think this is an oversight. As I mentioned in the comment #23, we should have moved the last patch to a new bug and fixed it there. In fact, I think that is what has happened. It is just that the reviewed patch here added confusion. Let me obsolete it and reopen the bug.

[Pavel Feldman]

Reopening the original issue.

[Timothy Hatcher]

We show both sizes in the Network panel if you have large rows enabled.