Backspace deletes all input

[Ondřej Surý]

Binary package hint: gcalctool

This is regression from Ubuntu Karmic.

gcalctool started to delete all entered input when backspace is pressed, e.g. if I type 2+1+34+2+1 and press BKSP I get empty input

When downgraded to 5.28.2-0ubuntu1 it deletes just last entered digit, e.g. if I type 2+1+34+2+1 and press BKSP I get "2+1+34+2+".

It's very annoying and makes the calculator almost unusable :-(.

ProblemType: Bug
Architecture: i386
Date: Fri Feb 12 16:39:29 2010
DistroRelease: Ubuntu 10.04
EcryptfsInUse: Yes
Package: gcalctool 5.29.90-0ubuntu1
ProcVersionSignature: Ubuntu 2.6.32-13.18-generic
SourcePackage: gcalctool
Uname: Linux 2.6.32-13-generic i686

[Monkey (monkey-libre)]

I´ve confirmed this bug. I have this problem each time that I do Backspace. Thank You for making Ubuntu better.

[Robert Ancell (robert-ancell)]

I'm not reproducing this, I do the following:
1. Open GCalctool
2. Type: "2+1+34+2+1"
display shows "2+1+34+2+1"
3. Press backspace
display show "2+1+34+2+"

[Robert Ancell (robert-ancell)]

What language/keyboard layout are you using?

[Monkey (monkey-libre)]

I´m using Spanish but before I used English.

[Ondřej Surý (ondrej)]

kbmap used is: us and cz

locale:
LANG=en_US.UTF-8
LANGUAGE=en_US.UTF-8
LC_CTYPE=cs_CZ.UTF-8
LC_NUMERIC=cs_CZ.UTF-8
LC_TIME=cs_CZ.UTF-8
LC_COLLATE=cs_CZ.UTF-8
LC_MONETARY=cs_CZ.UTF-8
LC_MESSAGES="en_US.UTF-8"
LC_PAPER=cs_CZ.UTF-8
LC_NAME=cs_CZ.UTF-8
LC_ADDRESS=cs_CZ.UTF-8
LC_TELEPHONE=cs_CZ.UTF-8
LC_MEASUREMENT=cs_CZ.UTF-8
LC_IDENTIFICATION=cs_CZ.UTF-8
LC_ALL=

But I get same behaviour when I unset all locale variables.

Are you sure that you have tried 5.29.90-0ubuntu1 (version from lucid)?

[Robert Ancell (robert-ancell)]

Using 5.29.90-0ubuntu1 on latest Lucid, did a fresh install and see the same behaviour. If you guys have some programming experience please download from source and try poking around in src/gtk.c main_window_key_press_cb() and see if you can catch anything going wrong.

[Robert Ancell (robert-ancell)]

(by same behaviour I mean it's still working correctly for me)

[Nicolò Chieffo (yelo3)]

I'm experiencing the same bug.
I have a very similar locale as Ondřej Surý, that is LANG,LANGUAGE and LC_MESSAGES set to en_US.UTF-8, but others set to my locale it_IT.UTF-8.

Robert Ancell, can you please see if changing the locale as we did triggers the bug?

[Robert Ancell (robert-ancell)]

I'm playing around with locales and keyboard layouts but still not reproducing.

[Nicolò Chieffo (yelo3)]

try to put this in ~/.profile

unset LC_ALL
export LANGUAGE="en_US.UTF-8"
export LANG="$LANGUAGE"
export LC_CTYPE="it_IT.UTF-8"
export LC_NUMERIC="it_IT.UTF-8"
export LC_TIME="it_IT.UTF-8"
export LC_COLLATE="it_IT.UTF-8"
export LC_MONETARY="it_IT.UTF-8"
export LC_PAPER="it_IT.UTF-8"
export LC_NAME="it_IT.UTF-8"
export LC_ADDRESS="it_IT.UTF-8"
export LC_TELEPHONE="it_IT.UTF-8"
export LC_MEASUREMENT="it_IT.UTF-8"
export LC_IDENTIFICATION="it_IT.UTF-8"

[Ondřej Surý (ondrej)]

I don't think it's locale related.

More like xkbmap related.

If you diff src/gtk.c, you'll find out they have added new keyboard shortcuts CTRL-BKSP and SHIFT-DEL to clear whole display. But apparently something is wrong in that function.

Ondrej

[Nicolò Chieffo (yelo3)]

If so, I have set the keymap to italian, can you try if this triggers the bug?

[Ondřej Surý (ondrej)]

I have traced it down to display_insert(), so after all it is locale related, see attached patch:

diff -urNap gcalctool-5.29.90~/src/display.c gcalctool-5.29.90/src/display.c
--- gcalctool-5.29.90~/src/display.c 2010-02-09 06:14:11.000000000 +0100
+++ gcalctool-5.29.90/src/display.c 2010-02-19 11:46:38.342465011 +0100
@@ -445,7 +445,7 @@ display_insert(GCDisplay *display, int c
                 use = FALSE;

             /* Ignore thousands separators */
- if (strncmp(c, v->tsep, strlen(v->tsep)) == 0)
+ if (*(v->tsep) && strncmp(c, v->tsep, strlen(v->tsep)) == 0)
                 use = FALSE;

             /* Copy existing text */

[Ondřej Surý (ondrej)]

BTW you are stripping all thousands separators when cursor_start isn't -1 in display_insert() call.

[Ondřej Surý (ondrej)]

This bug was reintroduced in 0be7b51e, so maybe better fix is to revert that small change, so it does adhere to your coding standards:

@@ -430,7 +473,7 @@ display_insert(GCDisplay *display, int cursor_start, int cursor_end, const char
                 use = FALSE;

             /* Ignore thousands separators */
- if (strncmp(c, v->tsep, strlen(v->tsep)) == 0)
+ if (v->tsep[0] != '\0' && strncmp(c, v->tsep, strlen(v->tsep)) == 0)
                 use = FALSE;

             /* Copy existing text */

[Robert Ancell (robert-ancell)]

Thanks Ondřej, fixed in 5.29.91
To reproduce:
1. LC_NUMERIC=it_IT.UTF-8 gcalctool
2. Enter "1234"
3. Press backspace
Expected:
"123" is displayed
Observed:
"" is displayed

[Launchpad Janitor (janitor)]

This bug was fixed in the package gcalctool - 5.29.91-0ubuntu1

---------------
gcalctool (5.29.91-0ubuntu1) lucid; urgency=low

  * New upstream release:
    * Fix bug where backspace deletes all input in locales with no thousands
      separator (LP: #521038)
    * Make boolean operations work with zero, e.g. "0 XOR 5" (LP: #496235)
    * Don't override space and return when display area not selected
    * Make window non-resizable (LP: #486359) (LP: #501680)
    * Change boolean functions to uppercase (xor -> XOR)
    * Documentation rewritten (LP: #521396)
    * Modernised build system
  * debian/control:
    - Build-depend on gnome-common
    - Remove build-depend on scrollkeeper
  * debian/rules:
    - Run autoreconf on build
    - Don't need to clean /var anymore (no scrollkeeper)
    - Don't need to disable scrollkeeper on build
    - Don't need to delete libparser.a anymore
  * debian/install:
  * debian/links:
  * debian/rules:
    - Install gcalctool.xpm from debian/install
    - Make gnome-calculator man page link in debian/links
  * debian/patches/01_lpi.patch:
    - Updated
  * debian/patches/99_autoconf.patch:
    - Not required, autoconf now run from debian/rules
 -- Robert Ancell <email address hidden> Mon, 22 Feb 2010 11:22:07 +1100