Echo / message error on load failures

[Ambrevar]

Recipe:

• disconnect from the Internet.
• load a page.

We need to tell the user that the page could not be loaded instead of doing nothing, since it's very confusing.

Ideally, the status view should keep the load status displayed since the echoed message is ephemeral.

[jmercouris]

Interesting, mine shows an error, says page could not be loaded after some time

[jmercouris]

Status view is now updated to show when a page is loading. The renderer will try as long as possible, and eventually report an error message. There are no events that should overwrite the status until it is done.

[jmercouris]

Works as expected:

[jmercouris]

I find it hilarious that my interface is half in English and half in Greek :-D

[Ambrevar]

Thanks, that's already an improvement! We can do a little bit better I think, consider this: Run `follow-hint-new-buffer` and follow a hint in the background. Notice that both the status view and the echo area report that they are loading the buffer. This is redundant, plus the current buffer does not display any meaningful status anymore. I suggest the following approach instead: when the buffer loads, have print-status display the regular status bar with (loading...) appended before or after the URL. This way it's contextual to the buffer that's actually loading. Thoughts?

[jmercouris]

Ah, I see what is happening, that behavior with the background was not intended... :-D

[jmercouris]

I don't understand your proposed solution, can you explain how that would fix the problem? The real issue is that the status message is currently tied to a window, and not a buffer

[Ambrevar]

My suggestion: - Don't use print-status like you did on a buffer load. - Instead, in the print-status function, get the load state of the buffer and print it next to the URL (we might need a new buffer slot for that). By the way, I'm realizing that the current fix is going off topic: the issue is about load _failures_, e.g. when there is not internet connection. Right now we get buffers like "Internel WebKit error" which means little to the user. What we must do here is handle the load-failed signal.

[jmercouris]

print-status would still need to be informed by someone, that would have to be on buffer load. are you suggesting the following:

print-status invoked in buffer load with no arguments
print status default status formatter adds loading if a buffer is currently loading?

[jmercouris]

By the way, I'm realizing that the current fix is going off topic: the issue is about load failures, e.g. when there is not internet connection. Right now we get buffers like "Internel WebKit error" which means little to the user. What we must do here is handle the load-failed signal.

Strange that it tells you "Internal Webkit Error". For me it says the service or servername could not be resolved (which is what is happening).

[Ambrevar]

print-status would still need to be informed by someone, that would have to be on buffer load. are you suggesting the following: print-status invoked in buffer load with no arguments print status default status formatter adds loading if a buffer is currently loading?

Yes!

[Ambrevar]

Strange that it tells you "Internal Webkit Error". For me it says the service or servername could not be resolved (which is what is happening).

Indeed, I never had this message myself.

[jmercouris]

The following code:

  (gobject:g-signal-connect
   (gtk-object buffer) "load-failed"
   (lambda (web-view load-event failing-uri error)
     (declare (ignore load-event failing-uri web-view error))
     (print "Load Fail")))

Produces an error:

Unknown GBoxed type 'GError'
   [Condition of type SIMPLE-ERROR]

Restarts:
 0: [ABORT] Exit debugger, returning to top level.

Backtrace:
  0: (GOBJECT::GET-G-BOXED-FOREIGN-INFO-FOR-GTYPE #<GTYPE :name "GError" :id 140228195051424>)
  1: ((:METHOD GOBJECT::PARSE-G-VALUE-FOR-TYPE (T (EQL (GOBJECT:GTYPE "GBoxed")))) #.(SB-SYS:INT-SAP #X01F38E68) #<unused argument>) [fast-method]
  2: ((FLET CALL-NEXT-METHOD :IN "/Users/jmercouris/.quicklisp/dists/quicklisp/software/cl-cffi-gtk-20200427-git/gobject/gobject.g-value.lisp"))
  3: ((:METHOD GOBJECT::PARSE-G-VALUE-FOR-TYPE :AROUND (T T)) #.(SB-SYS:INT-SAP #X01F38E68) #<GTYPE :name "GBoxed" :id 72>) [fast-method]
  4: ((SB-PCL::GF-DISPATCH GOBJECT::PARSE-G-VALUE-FOR-TYPE) #.(SB-SYS:INT-SAP #X01F38E68) #<GTYPE :name "GBoxed" :id 72>)
  5: ((:METHOD GOBJECT::PARSE-G-VALUE-FOR-TYPE (T T)) #.(SB-SYS:INT-SAP #X01F38E68) #<GTYPE :name "GError" :id 140228195051424>) [fast-method]
  6: ((FLET CALL-NEXT-METHOD :IN "/Users/jmercouris/.quicklisp/dists/quicklisp/software/cl-cffi-gtk-20200427-git/gobject/gobject.g-value.lisp"))
  7: ((:METHOD GOBJECT::PARSE-G-VALUE-FOR-TYPE :AROUND (T T)) #.(SB-SYS:INT-SAP #X01F38E68) #<GTYPE :name "GError" :id 140228195051424>) [fast-method]
  8: ((SB-PCL::GF-DISPATCH GOBJECT::PARSE-G-VALUE-FOR-TYPE) #.(SB-SYS:INT-SAP #X01F38E68) #<GTYPE :name "GError" :id 140228195051424>)
  9: (GOBJECT:PARSE-G-VALUE #.(SB-SYS:INT-SAP #X01F38E68))
 10: (GOBJECT::PARSE-CLOSURE-ARGUMENTS 4 #.(SB-SYS:INT-SAP #X01F38E20))
 11: ((LAMBDA (GOBJECT::CLOSURE GOBJECT::RETURN-VALUE GOBJECT::COUNT-OF-ARGS GOBJECT::ARGS GOBJECT::INVOCATION-HINT GOBJECT::MARSHAL-DATA) :IN "/Users/jmercouris/.quicklisp/dists/quicklisp/software/cl-cffi..
 12: ((LAMBDA (SB-ALIEN::ARGS-POINTER SB-ALIEN::RESULT-POINTER FUNCTION) :IN "/Users/jmercouris/.quicklisp/dists/quicklisp/software/cl-cffi-gtk-20200427-git/gobject/gobject.signals.lisp") #<unavailable arg..
 13: ("foreign function: funcall_alien_callback")
 14: ("foreign function: #x2010073B")
 15: ("foreign function: #x1CB7E00")
 16: ("foreign function: signal_emit_unlocked_R")
 17: ("foreign function: g_signal_emit_valist")
 18: ("foreign function: g_signal_emit")
 19: ("foreign function: _Z23webkitWebViewLoadFailedP14_WebKitWebView15WebKitLoadEventPKcP7_GError")
 --more--

here is the relevant section of the API:
https://webkitgtk.org/reference/webkit2gtk/stable/WebKitWebView.html#WebKitWebView-load-failed

I believe this is an error in cl-cffi-gtk, I do think we can handle it in cl-webkit

ideas?

[Ambrevar]

I can reproduce. Can you report to https://github.com/Ferada/cl-cffi-gtk/ ?

[jmercouris]

I will try to report, and I will try the other cl-cffi-gtk as well, I want to do some more investigation first though

[jmercouris]

BAM!

[jmercouris]

[Ambrevar]

Fantastic!

Note for posterity:

• please mention the commit where this is fixed (here ab813b3).
• When you upload a submodule, let me know so that I don't forget to update the cl-webkit Guix package.

[jmercouris]

OK! I will try to do that in the future!

[jmercouris]

Fixed here: 9a96ff1