Bug 116225 – [GTK] Remote Web Inspector sometimes won't serve any content

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Simon Pena]

Accessing the remote web inspector sometimes results in an empty page, as it won't serve any content.

With WebKit built but not installed, and the environment variable WEBKIT_INSPECTOR_SERVER pointing to a valid address (e.g. 127.0.0.1:2999), launch MiniBrowser and then try to access the access given from a different browser.

In my research, this is caused by the WebInspectorServerGtk not finding the needed resources in the expected path. It could probably be fixed by providing an environment variable (in the same fashion done in the InspectorClientGtk) so users can provide their own inspector resources. (Other, maybe more sophisticated, methods could try to guess where this content is stored).

[Anton Obzhirov]

OK, I'll fix it. I think someone complained about it last year as well.

[Anton Obzhirov]

Created attachment 201967 [details]
Patch

[Simon Pena]

CCing Gustavo, who reviewed the initial implementation of GTK's remote web inspector in #88094

[Gustavo Noronha (kov)]

Comment on attachment 201967 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=201967&action=review

> Source/WebKit2/ChangeLog:3
> +        [GTK] Remote Web Inspector sometimes won't serve any content

I feel this bug title is a bit misleading, perhaps it should say that the remote inspector is only able to serve files from the configured install path.

[Gustavo Noronha (kov)]

Comment on attachment 201967 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=201967&action=review

> Source/WebKit2/UIProcess/InspectorServer/WebInspectorServer.h:76
> +    GOwnPtr<gchar> m_inspectorServerFilesPath;

Oh, actually, why not just use CString?

[Anton Obzhirov]

Well, in this case I just wanted smart pointer to own char string. Didn't really need other functionality of CString. Seems to be safer to use GOwnPtr. I think in similar code in InspectorClientGtk they use smart pointer as well.

[Gustavo Noronha (kov)]

We used char pointers a lot in the early days of webkitgtk, but we have mostly switched to using CString for storing strings in more recent times to avoid allocation churn and make the code simpler. I don't feel too strongly about this, but I think using a smart pointer is unnecessary in this case.

[Anton Obzhirov]

ok, point taken. I will resubmit the patch.

[Anton Obzhirov]

Created attachment 202438 [details]
Patch

[Gustavo Noronha (kov)]

Comment on attachment 202438 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=202438&action=review

> Source/WebKit2/UIProcess/InspectorServer/WebInspectorServer.h:75
> +    const char* inspectorServerFilesPath();

I think it doesn't make sense to use a char pointer just here, just return the CString? In fact, you can just use WTFString and avoid the other include IMO.

[Anton Obzhirov]

Created attachment 202514 [details]
Patch

[Anton Obzhirov]

I used webkit-patch apply-from-bug script option. So it still shows Gustavo as a reviewer.

[Gustavo Noronha (kov)]

Comment on attachment 202514 [details]
Patch

Thanks!

[WebKit Commit Bot]

Comment on attachment 202514 [details]
Patch

Clearing flags on attachment: 202514

Committed r150582: <http://trac.webkit.org/changeset/150582>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.