46992 - ctrl-backspace and ctrl-delete do not send correct keycode

Status: VERIFIED FIXED

(Core :: DOM: UI Events & Focus Handling, defect, P3)

Description

[Kathleen :Brade]

ctrl-backspace and ctrl-delete do not send correct keycode
While trying to fix bug #41078, I found that control-delete doesn't send 
VK_DELETE to the xul keylistener, instead the OS is sending a capital Beta 
character (Greek symbol; 233 decimal if I remember correctly).  For Control-
Backspace, the character 'h' is sent.  The proper keycodes *are* sent if the 
control key is not pressed.

NOTE:  This is a bug in the OS possibly??

Comment 1

[Peter Trudelle]

need info: Is this still reproducible?  What are those keys needed for?  Is
there any workaround?

Comment 2

[Peter Trudelle]

nsbeta3-, ->future.  Please renominate with more info if you don't agree.

Comment 3

[Dan Nunn]

Not sure if this is related, but win98 20000818:

Pressing shift or ctrl while pressing backspace or delete causes nothing to
happen, whereas in other applications, Windows quietly ignores the shift or
control and backspaces or deletes appropriately. Why is this a big deal? For
those of us like me who suck at typing, and sometimes leave their fingers on
Shift when hitting backspace, and it doesn't do anything. If this is completely
unrelated to the original reporter's bug, tell me and I'll quietly go away (and
file a new bug).

Comment 4

[WD]

I too have noticed the Shift + BS problem on my system.    Not sure if it is a 
separate bug from the original one filed here, though.

Let's say I make a mistake while typing something in ALL CAPZ^HS
I'm not going to let go of the shift key just to hit the backspace.   With 
mozilla, the backspace is useless when used in this manner.

Perhaps a new bug should be filed that explicitly describes this problem.

Comment 5

[Peter Trudelle]

It sounds like what you're describing are scenarios that would be caused by the
originally described defect.  I still don't think this makes our short list for
nsbeta3, but we'd consider a good patch.    ->helpwanted.

Comment 6

[Akkana Peck]

You might want to take a look at my patch for bug 50252.  The issue in that bug
turned out to be that gtk, for some reason, when you get a control-[alpha], sets
the string to the ordinal number of the character rather than the character
itself, e.g. for ctrl-A you get \001 rather than a, and so there was code (which
was always being applied) to add 'a' - 1 or 'A' - 1.  I fixed 50252 with a patch
that says if isprint() and isalpha(), then return the character, otherwise do
the +'a'-1 thing.  The isprint() means it won't fix this bug directly -- you
still need to find a way to distinguish the codes you want to send directly, and
of course, backspace will send a ctrl-h (\008) whereas h will send 'h' - 'a' + -
(\008).  So I don't see how we'll be able to tell them apart.

This seems like really broken behavior on gtk's part.  Pav, any idea why it does
this and whether there's another way to tell them apart?

Comment 7

[WD]

Just FYI, my Shift+BS problem is on a Win32 system.

Shift+Backspace on mozilla doesn't work.    With IE and Netscape, no problem.

Comment 8

[Kathleen :Brade]

sorry I haven't been able to respond sooner, I've been away on vacation.

This is a GTK-specific bug.

The windows problem described should be filed as a new bug (please cc me on that 
bug).

Comment 9

[Akkana Peck]

I just checked in a fix for bug 50252 which should fix this as well (on Unix; as
Kathy said, any problems on Windows would be a separate bug).  Marking fixed.

Comment 10

[sairuh (rarely reading bugmail)]

using 2000.10.03.xx-n6 [opt commercial branch bits], shift+backspace works (ie,
just like backspace). however, ctrl+backspace, ctrl+delete and shift+delete do
nothing.

akkana, what in particular was fixed here? or did i just stumble into a
regression (and thus should reopen this)?

Comment 11

[Akkana Peck]

What was fixed was the keycodes they sent -- so if there are any bindings on
those keycodes, they wouldn't have worked.

I'm not sure of whether we actually have bindings to these keys in the current
key binding list (the un-modified keys should work, of course, but they're
handled in the editor via a direct event handler, not via xul or xbl bindings)
so I'm not sure whether this is testable or not, without explicitly setting up
bindings to e.g. ctrl-delete and seeing if they work.

Comment 12

[sairuh (rarely reading bugmail)]

hm, okay, rubberstamp vrfy.