[PATCH 0/2] vt: mouse selection word boundaries

[PATCH 0/2] vt: mouse selection word boundaries

[Adam Borowski]

Hi!
Here's a couple of really low priority fixes to how "word chars" for mouse
selection are determined.

Patch 1 (adds "-./") is an epitome of "apply if bored": for two decades,
only gpm used this, and it always ignored the defaults.  Bill Allombert made
a second implementation, "consolation", and we wasted a bit of time figuring
out why mouse selection is so limited.  As consolation now overrides the
defaults too, this patch will be helpful only for whoever makes a third user
of this code...

Patch 2, besides making a handful of non-ASCII symbols work the same as
everything else, also stops people from mysteriously losing that 'ü' after
naive gpm -l "-A-Za-z0-9_./".

-- 
⢀⣴⠾⠻⢶⣦⠀ Meow!
⣾⠁⢠⠒⠀⣿⡁
⢿⡄⠘⠷⠚⠋⠀ Collisions shmolisions, let's see them find a collision or second
⠈⠳⣄⠀⠀⠀⠀ preimage for double rot13!

[PATCH 1/2] vt: set mouse selection word-chars to gpm's default

[Adam Borowski]

Since forever, gpm was this code's only user, and it overrides the table on
start so the default was never seen -- until Bill Allombert's "consolation"
came in.  The in-kernel set is "A-Za-z0-9_" which fails to catch typical
file names, etc.  Let's change this to gpm's conservative default, ie
"-A-Za-z0-9_./"; most terminals include more, for example xfce4-terminal has
"-A-Za-z0-9,./?%&#:_=+@~".

There's some discussion at https://bugs.debian.org/846587

Signed-off-by: Adam Borowski <kilobyte@angband.pl>
---
 drivers/tty/vt/selection.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/tty/vt/selection.c b/drivers/tty/vt/selection.c
index 36e1b8c7680f..2252e11d8347 100644
--- a/drivers/tty/vt/selection.c
+++ b/drivers/tty/vt/selection.c
@@ -84,7 +84,7 @@ void clear_selection(void)
  */
 static u32 inwordLut[8]={
   0x00000000, /* control chars     */
-  0x03FF0000, /* digits            */
+  0x03FFE000, /* digits and "-./"  */
   0x87FFFFFE, /* uppercase and '_' */
   0x07FFFFFE, /* lowercase         */
   0x00000000,
-- 
2.11.0

[PATCH 2/2] vt: make mouse selection of non-ASCII consistent

[Adam Borowski]

For some reason a handful of ISO-8859-1 symbols are excluded from "word
chars" while the vast majority of Unicode is hard-coded as included, even
when inappropriate (we really would want to _not_ select line-drawing/etc).
Those symbols are: ¡¢£¤¥¦§¨©ª«¬®¯°±²³´µ¶·¸¹º»¼½¾¿×÷

Thus, let's not special-case any non-ASCII anymore.  Attempts to set these
via ioctl will be silently ignored.

As an extra bonus, we debloat the kernel by 128 bytes.

Signed-off-by: Adam Borowski <kilobyte@angband.pl>
---
 drivers/tty/vt/selection.c | 16 ++++++----------
 1 file changed, 6 insertions(+), 10 deletions(-)

diff --git a/drivers/tty/vt/selection.c b/drivers/tty/vt/selection.c
index 2252e11d8347..c81c99165ea6 100644
--- a/drivers/tty/vt/selection.c
+++ b/drivers/tty/vt/selection.c
@@ -80,21 +80,17 @@ void clear_selection(void)
 
 /*
  * User settable table: what characters are to be considered alphabetic?
- * 256 bits. Locked by the console lock.
+ * 128 bits. Locked by the console lock.
  */
-static u32 inwordLut[8]={
+static u32 inwordLut[4]={
   0x00000000, /* control chars     */
   0x03FFE000, /* digits and "-./"  */
   0x87FFFFFE, /* uppercase and '_' */
   0x07FFFFFE, /* lowercase         */
-  0x00000000,
-  0x00000000,
-  0xFF7FFFFF, /* latin-1 accented letters, not multiplication sign */
-  0xFF7FFFFF  /* latin-1 accented letters, not division sign */
 };
 
 static inline int inword(const u16 c) {
-	return c > 0xff || (( inwordLut[c>>5] >> (c & 0x1F) ) & 1);
+	return c > 0x7f || (( inwordLut[c>>5] >> (c & 0x1F) ) & 1);
 }
 
 /**
@@ -106,10 +102,10 @@ static inline int inword(const u16 c) {
  */
 int sel_loadlut(char __user *p)
 {
-	u32 tmplut[8];
-	if (copy_from_user(tmplut, (u32 __user *)(p+4), 32))
+	u32 tmplut[4];
+	if (copy_from_user(tmplut, (u32 __user *)(p+4), 16))
 		return -EFAULT;
-	memcpy(inwordLut, tmplut, 32);
+	memcpy(inwordLut, tmplut, 16);
 	return 0;
 }
 
-- 
2.11.0

Re: [PATCH 2/2] vt: make mouse selection of non-ASCII consistent

[Jiri Slaby]

On 03/27/2017, 12:37 PM, Adam Borowski wrote:
> For some reason a handful of ISO-8859-1 symbols are excluded from "word
> chars" while the vast majority of Unicode is hard-coded as included, even
> when inappropriate (we really would want to _not_ select line-drawing/etc).
> Those symbols are: ¡¢£¤¥¦§¨©ª«¬®¯°±²³´µ¶·¸¹º»¼½¾¿×÷
> 
> Thus, let's not special-case any non-ASCII anymore.  Attempts to set these
> via ioctl will be silently ignored.
> 
> As an extra bonus, we debloat the kernel by 128 bytes.
> 
> Signed-off-by: Adam Borowski <kilobyte@angband.pl>
> ---
>  drivers/tty/vt/selection.c | 16 ++++++----------
>  1 file changed, 6 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/tty/vt/selection.c b/drivers/tty/vt/selection.c
> index 2252e11d8347..c81c99165ea6 100644
> --- a/drivers/tty/vt/selection.c
> +++ b/drivers/tty/vt/selection.c
> @@ -80,21 +80,17 @@ void clear_selection(void)
>  
>  /*
>   * User settable table: what characters are to be considered alphabetic?
> - * 256 bits. Locked by the console lock.
> + * 128 bits. Locked by the console lock.
>   */
> -static u32 inwordLut[8]={
> +static u32 inwordLut[4]={

No need for the constant here. Let it autosize.

>    0x00000000, /* control chars     */
>    0x03FFE000, /* digits and "-./"  */
>    0x87FFFFFE, /* uppercase and '_' */
>    0x07FFFFFE, /* lowercase         */
> -  0x00000000,
> -  0x00000000,
> -  0xFF7FFFFF, /* latin-1 accented letters, not multiplication sign */
> -  0xFF7FFFFF  /* latin-1 accented letters, not division sign */
>  };
>  
>  static inline int inword(const u16 c) {
> -	return c > 0xff || (( inwordLut[c>>5] >> (c & 0x1F) ) & 1);
> +	return c > 0x7f || (( inwordLut[c>>5] >> (c & 0x1F) ) & 1);
>  }
>  
>  /**
> @@ -106,10 +102,10 @@ static inline int inword(const u16 c) {
>   */
>  int sel_loadlut(char __user *p)
>  {
> -	u32 tmplut[8];
> -	if (copy_from_user(tmplut, (u32 __user *)(p+4), 32))
> +	u32 tmplut[4];

ARRAY_SIZE(inwordLut) here.

> +	if (copy_from_user(tmplut, (u32 __user *)(p+4), 16))
>  		return -EFAULT;
> -	memcpy(inwordLut, tmplut, 32);
> +	memcpy(inwordLut, tmplut, 16);

sizeof(inwordLut) here and for copy_from_user too.

>  	return 0;
>  }

thanks,
-- 
js
suse labs

[PATCH v2 1/2] vt: set mouse selection word-chars to gpm's default

[Adam Borowski]

Since forever, gpm was this code's only user, and it overrides the table on
start so the default was never seen -- until Bill Allombert's "consolation"
came in.  The in-kernel set is "A-Za-z0-9_" which fails to catch typical
file names, etc.  Let's change this to gpm's conservative default, ie
"-A-Za-z0-9_./"; most terminals include more, for example xfce4-terminal has
"-A-Za-z0-9,./?%&#:_=+@~".

There's some discussion at https://bugs.debian.org/846587

Signed-off-by: Adam Borowski <kilobyte@angband.pl>
---
v2: no changes in this patch, resent as part of the set.

 drivers/tty/vt/selection.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/tty/vt/selection.c b/drivers/tty/vt/selection.c
index 36e1b8c7680f..2252e11d8347 100644
--- a/drivers/tty/vt/selection.c
+++ b/drivers/tty/vt/selection.c
@@ -84,7 +84,7 @@ void clear_selection(void)
  */
 static u32 inwordLut[8]={
   0x00000000, /* control chars     */
-  0x03FF0000, /* digits            */
+  0x03FFE000, /* digits and "-./"  */
   0x87FFFFFE, /* uppercase and '_' */
   0x07FFFFFE, /* lowercase         */
   0x00000000,
-- 
2.11.0

[PATCH v2 2/2] vt: make mouse selection of non-ASCII consistent

[Adam Borowski]

For some reason a handful of ISO-8859-1 symbols are excluded from "word
chars" while the vast majority of Unicode is hard-coded as included, even
when inappropriate (we really would want to _not_ select line-drawing/etc).
Those symbols are: ¡¢£¤¥¦§¨©ª«¬®¯°±²³´µ¶·¸¹º»¼½¾¿×÷

Thus, let's not special-case any non-ASCII anymore.  Attempts to set these
via ioctl will be silently ignored.

As an extra bonus, we debloat the kernel by 128 bytes.

Signed-off-by: Adam Borowski <kilobyte@angband.pl>
---
v2: Got rid of hard-coded array sizes.

 drivers/tty/vt/selection.c | 16 ++++++----------
 1 file changed, 6 insertions(+), 10 deletions(-)

diff --git a/drivers/tty/vt/selection.c b/drivers/tty/vt/selection.c
index 2252e11d8347..accbd1257bc4 100644
--- a/drivers/tty/vt/selection.c
+++ b/drivers/tty/vt/selection.c
@@ -80,21 +80,17 @@ void clear_selection(void)
 
 /*
  * User settable table: what characters are to be considered alphabetic?
- * 256 bits. Locked by the console lock.
+ * 128 bits. Locked by the console lock.
  */
-static u32 inwordLut[8]={
+static u32 inwordLut[]={
   0x00000000, /* control chars     */
   0x03FFE000, /* digits and "-./"  */
   0x87FFFFFE, /* uppercase and '_' */
   0x07FFFFFE, /* lowercase         */
-  0x00000000,
-  0x00000000,
-  0xFF7FFFFF, /* latin-1 accented letters, not multiplication sign */
-  0xFF7FFFFF  /* latin-1 accented letters, not division sign */
 };
 
 static inline int inword(const u16 c) {
-	return c > 0xff || (( inwordLut[c>>5] >> (c & 0x1F) ) & 1);
+	return c > 0x7f || (( inwordLut[c>>5] >> (c & 0x1F) ) & 1);
 }
 
 /**
@@ -106,10 +102,10 @@ static inline int inword(const u16 c) {
  */
 int sel_loadlut(char __user *p)
 {
-	u32 tmplut[8];
-	if (copy_from_user(tmplut, (u32 __user *)(p+4), 32))
+	u32 tmplut[ARRAY_SIZE(inwordLut)];
+	if (copy_from_user(tmplut, (u32 __user *)(p+4), sizeof(inwordLut)))
 		return -EFAULT;
-	memcpy(inwordLut, tmplut, 32);
+	memcpy(inwordLut, tmplut, sizeof(inwordLut));
 	return 0;
 }
 
-- 
2.11.0