Clipboard is unreliable with thunderbird editor area

[psycho_zs]

Running thunderbird on server, copying or pasting text from/into thunderbird's editor area not working most of the time.

In the test I've copied text from leafpad on the client to leafpad on server (successful), then another text from leafpad on the client to thunderbird on server (previous data is pasted). Then again (unsuccessful), then pasted to leafpad on server (successful), then pasted to thunderbird (after pasted to leafpad, successful).

So after using clipboard for the first time, I'm only being able to paste to thunderbird after I paste to something else.

[Antoine Martin]

From the log samples: xpra X11 version 2.0-r15223 64-bit on Linux Debian 9.0 stretch.

Oh god, another clipboard bug..

[Antoine Martin]

I cannot reproduce.
The clipboard works for me everytime, in both directions.
I've tried on Fedora 25 and Debian Stretch.

Maybe something is interfering with the clipboard on your client system? See wiki/Clipboard.

I cannot use the sample log files provided as I have no way of knowing which part of the log file relate to what: the communication is done via the X11 server's "selections", so all the events look pretty much the same.

[Antoine Martin]

Please provide more succinct or annotated log samples, otherwise I will have to close this ticket as "worksforme".

[psycho_zs]

I've isolated two events: I copy text from leafpad on client and unsuccessfully try to paste it into thunderbird on server. No clipboard manager running. Copied text was: 2:36: copy from leafpad client to tb server
Attaching this part of server log and client output...

[psycho_zs]

isolated copy+paste

[psycho_zs]

isolated copy+paste

[Antoine Martin]

OK, so that's quite a bit clearer.
We see lots of:

process clipboard request, request_id=11, selection=CLIPBOARD, local name=CLIPBOARD, target=TARGETS

That's the application requesting the list of targets for the clipboard. (the list of formats we can supply for this clipboard)

We dutifully respond every time with the list:

['TIMESTAMP', 'TARGETS', 'MULTIPLE', 'GTK_TEXT_BUFFER_CONTENTS', \
    'application/x-gtk-text-buffer-rich-text', 'UTF8_STRING', 'TEXT', 'STRING',  \
    'text/plain;charset=utf-8', 'text/plain']

But then we get a request that doesn't make any sense:

process clipboard request, request_id=14, selection=CLIPBOARD, local name=CLIPBOARD, target=application/x-moz-nativehtml

As it uses a target ("application/x-moz-nativehtml") we have not advertised.
And when forward it to the application, it responds with "no data":

unpack(..) type=, format=0, data=<type 'NoneType'>:0

Which is expected since we're requesting a target it knows nothing about.

And this rings a bell:

• ticket:883#comment:11 : firefox also requesting "x-moz-nativehtml" when it shouldn't
• ticket:834#comment:6 : same issue, contains a hack-nativehtml-to-string.patch which can substitute the target with one that is actually going to work

The patch above was not applicable as it is: some applications may provide the "x-moz-nativehtml" target, so we should only translate this target if it isn't going to work.
r15589 does that, it also exposes the targets we have last sent via xpra info:

$ xpra info | grep last-targets

@psycho_zs: does that work for you now?

[Antoine Martin]

Not heard back, closing.