Re: Poor performance on a slow link
Unfortunately that wasn't quite enough. That patch is basically ok (but
see below), but when hack_motif_clipboard_selection is called owned_p is
always zero which makes the new x_selection_strict_motif_ownership switch
irrelevant. The problem is that own-selection-internal is always called
(as far as I can tell) with the how-to-add parameter defaulted to nil. In
this case it doesn't set prev_value and hence owned_p is nil when the
device method is called. The attached patch seem to fix this.
It also seems to me that the documentation for
x-selection-strict-motif-ownership is backwards. It says that if it is
true (the default) then XEmacs won't own the clipboard again if it already
owns it, but it seems to be just the opposite. Since it's not clear to me
which the default should be and whether the documentation or the code is
wrong, I didn't change anything.
Mike Alexander <mailto:mta＠arbortext.com>
Arbortext, Inc. +1-734-997-0200
--On Tuesday, January 23, 2001 7:51 PM +0000 Andy Piper <andy(a)xemacs.org>
wrote:
Try this.
andy
2001-01-23 Andy Piper <andy.piper(a)beasys.org>
* select.c (Fown_selection_internal): pass owned_p
* select-msw.c (mswindows_own_selection): New Signature.
* console.h (struct console_methods): add owned_p to
_own_selection.
* select-x.c (x_own_selection): pass owned_p
(hack_motif_clipboard_selection): use owned_p
(vars_of_select_x): new variable -
x_selection_strict_motif_ownership.
At 04:12 04/01/01 -0500, Mike Alexander wrote:
> I've noticed that both 21.2.38 and 21.2.39 have a problem when there is
> a slow link like an ISDN line between the X Server and a copy of XEmacs
> running on Solaris. Some operations (such as copy-primary-selection or
> kill-line) transmit a lot more data than they should, around 30K. Other
> similar operations (such as advertized-undo) are very fast and transmit
> almost no data. I usually work with the Windows version of XEmacs and
> don't know much about X or the use of it by XEmacs so I don't know
> what's going on. I can send someone a trace of the traffic, but it
> should be easy to reproduce since it happens even with when XEmacs is
> run with -vanilla. I don't know when this started happening since I
> haven't built a Unix version for a while. I'm also behind in my EMail
> so this may have been discussed although I didn't see it in a quick
> scan. If it's an old topic, I apologize.
>
> Mike Alexander <mailto:mta＠arbortext.com>
> Arbortext, Inc. +1-734-997-0200
Attachments:
- select.c.diff (application/octet-stream — 1.3 KB)