Re: [PATCH] [gdb/tui] Handle unicode chars in prompt

[Tom de Vries <tdevries@suse.de>]

On 5/31/23 13:29, Tom de Vries wrote:
> On 5/30/23 19:03, Tom Tromey wrote:
>>>> In TUI, the prompt is written out by tui_puts_internal, which 
>>>> outputs one byte
>>>> at a time using waddch, which apparantly breaks multi-byte char 
>>>> support.
>>>> Fix this by detecting multi-byte chars in tui_puts_internal, and
>>>> printing them using
>>>> waddnstr.
>>
>>> FWIW, I just came across this commit, which seems relevant:
>>
>> Tom> Note that tui_puts_internal remains.  It is needed to handle 
>> computing
>> Tom> the start line of the readline prompt, which is difficult to do
>> Tom> properly in the case where redisplaying can also cause the command
>> Tom> window to scroll.  This might be possible to implement by 
>> reverting to
>> Tom> single "character" output, by using mbsrtowcs for its side 
>> effects to
>> Tom> find character boundaries in the input.  I have not attempted this.
>> Tom> ...
>>
>> I no longer remember what made this difficult.  I wonder if it's
>> possible to simply emit as many characters as possible in a single call,
>> and then use getyx to figure out the length of the prompt after it has
>> been fully displayed.  If the prompt wraps or if it takes multiple
>> lines, offhand it seems fine to just pick whatever the final column
>> happens to be.
> 
> I've given that a try, and that seems to work.
> 
> I also realized that we don't cover wrapping prompts in the testsuite, 
> so I wrote a test-case ( 
> https://sourceware.org/pipermail/gdb-patches/2023-May/199950.html ).

I've committed the test-case (excluding the proposed fix for now), with 
an extra check that FAILs in combination with this patch:
...
FAIL: gdb.tui/long-prompt.exp: prompt size == width + 1: end of screen: 
scrolling
...
because we have:
...
    17 (gdb) set prompt 123456789A123456789B123
    18 456789C123456789D>
    19 123456789A123456789B123456789C123456789D
    20 123456789A123456789B123456789C123456789D
    21 123456789A123456789B123456789C123456789D
    22 >set prompt (gdb)
    23 (gdb)
...
instead of the expected:
...
    19 (gdb) set prompt 123456789A123456789B123
    20 456789C123456789D>
    21 123456789A123456789B123456789C123456789D
    22 >set prompt (gdb)
    23 (gdb)
...

The logic I used in this patch:
...
+  if (height != nullptr)
+    {
+      int line = getcury (w);
+      *height += line - prev_line;
      }
...
was to use the current line to detect a wrap but that doesn't work if 
writing the prompt wraps at the last line which then generates a scroll.

This is the reason that a wrap is detected in the original code using a 
reduction in wrap position.

But AFAIU that doesn't work either for this "maximum-string" approach.
If the string is long enough, it's possible to wrap and increase column 
position.

In conclusion, AFAIU this approach doesn't work.

Thanks,
- Tom