[webkit-gtk] webkit2gtk resources

[Victor Lucero]

El 28-08-2014, a las 12:20, Jose <jmalv04 at gmail.com> escribió:

> I think the api is general enough that should accommodate tool users
> (as it did in webkit1).
> 
> I think collecting "got-chunk" pieces and  assembling them again is
> replicating what the code is already doing.
> Did you get any suggestions on alternative ways to do this ?

just getting the resource as you say but i needed the complete network timing cycle so i ended hooking myself into the soup session for this. Now i got DNS resolve times, Latency times, SSL times and first byte times. none of these elements can be obtained from the current api :( (api 1 or 2)

> 
> Wrt frames, I am only looking for the raw frame html, as this is a
> subresource of the main_resource. And as such
> the raw frame html content could/should be exposed in the UI process
> (the subresource is there but the content
> is empty).

We landed in the same issue is because of that we used  the soup session approach, also comparing all the elements downloaded from the page and the resources. they don’t match.
the network engine almost always ended downloaded more things than “resource-start” events where launched.

if i put all the elements downloaded vs all the elements in the page they never match .

> 
> I understand the separation between the UI and the network/web process
> but the devil is in the details. Is there any
> sample code of a specific web extension doing something useful ?
> 
> 
> 
> 
> 
> On Thu, Aug 28, 2014 at 6:07 PM, Victor Lucero
> <victor.lucero at ipartner.cl> wrote:
>> 
>> Comments below
>> 
>> El 28-08-2014, a las 11:51, Jose <jmalv04 at gmail.com> escribió:
>> 
>>> thanks, comments below
>>> 
>>> On Thu, Aug 28, 2014 at 4:32 PM, Victor Lucero
>>> <victor.lucero at ipartner.cl> wrote:
>>>> 
>>>> 1)Got the actual data for all the resources been downloaded its a bit
>>>> tricky. in our case we use webkit_get_default_session() to access the soup
>>>> main session and attach a "request-queued” event into it so for each element
>>>> downloaded you could access the soup-message, and for each soup-message you
>>>> can attach "got-chunk” so every piece of data is send to that event and you
>>>> need to manually store it to get its length and contents-  sadly this events
>>>> was removed in libsoup 2.39.4 if I remember well. but thanks that its open
>>>> source I patched that feature back (ugly but I was desperate) if you want i
>>>> cane send you that patch , its works in the latest version of lib soup used
>>>> by webkitgtk and webkit2gtk.
>>>> Sadly there is no element in the api to get or catch the frames that are
>>>> created in a webpage, the guys here told me to file a bug asking for this
>>>> feature in webkit2gtk but in the old API is present via the signal
>>>> “frame-created” over a web_view or frame. I think i will file that bug
>>>> because i need that functionality but im still working with the old API.
>>> 
>>> This approach seems overly complicated but it may well be the only
>>> approach with webkit2. Please do send me
>>> the patch as it will help me learn a bit more about this.
>> 
>> isn’t thaaat complicated. just attach some signals to the soup engine and wait to the contents and store them for later use.
>> i attached the patch but remember this only bring back the “got-chunk” functionality to the soup-message object of lib soup. it has nothing to do directly with webkitgtk.
>> (the path its just the diff between the version of the file when the feature was removed so this code isn mine ok?)
>> 
>> 
>>> 
>>> I think we should get guidance from the webkit2gtk developers (most
>>> code I've looked at is written by Carlos G,
>>> and it's nicely crafted :))
>>> 
>>> I found that webkit1 had a wekit_web_view_get_subresources
>>> http://trac.webkit.org/changeset/143302
>>> but it was removed for webkit2. I think something like that with a
>>> custom timeout to limit the caching of resources
>>> might be useful but there might be a better approach.
>>> 
>>>> 2) the method and patch described in point 1 also works for all the elements
>>>> that pass throughout the browser so all the elementos downloaded by the
>>>> flash and the fly itself can be obtained :)
>>> 
>>> Aha! This is then very useful. But still I think getting the
>>> subresources (in my case I collect them
>>> manually, via resource-load-started) should also include the frame raw
>>> content. The subresource
>>> seems to be there and the content is missing (unless I am missing something).
>>> 
>>> Also, I wanted some guidance on how to approach this with the new plugin changes
>>> http://blogs.igalia.com/carlosgc/2014/08/06/gtk-3-plugins-in-webkitgtk-and-evince-browser-plugin/
>>> 
>>> I want to detect when a specific plugin loader is being called and the
>>> content file.
>>> 
>> 
>> 
>> 
>> 
>>>> 3) how are you capturing these screenshots? We do screenshots accessing
>>>> Cairo but our tool make sure that the page is full loaded, also found that
>>>> some pages just crash the xlib if the page is too heavy and you don’t use a
>>>> scrollview to contain it. please, tell me more details about this issue.
>>> 
>>> I do very basic stuff but it works nicely. Just check the testing code
>>> for how to use
>>> webkit_web_view_get_snaphsot. Crashing shouldn't be a problem with the
>>> new process model.
>>> 
>>> Ping me offline if you need some ideas.
>>> 
>>> 
>>>> I hope this help
>>> 
>>> yes definitely.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4157 bytes
Desc: not available
URL: <https://lists.webkit.org/pipermail/webkit-gtk/attachments/20140828/ea00c2f2/attachment.p7s>