Re: [PATCH 0/2] gdb/tui: Assembler window scrolling fixes

[Shahab Vahedi <Shahab.Vahedi@synopsys.com>]

Hello Andrew,

I have tested the patch series against the "hello world"
program that I have [1]. Here comes my observations:

1. As you can see in this gif file:
https://sourceware.org/bugzilla/attachment.cgi?id=12200
"Going up" sometimes does not work. In this case, it is
between "main" and "_start" (garbage?). Nevertheless, I
manged to have it working for me by:

   @@ -161,7 +161,7 @@ tui_find_symbol_backward (CORE_ADDR addr)
 {
   struct bound_minimal_symbol msym;
 
-  for (int offset = 1; offset <= 1024; offset *= 2)
+  for (int offset = 1; offset <= 1024; ++offset)
     {
       CORE_ADDR tmp = addr - offset;
       msym = lookup_minimal_symbol_by_pc_section (tmp, 0);

2. I run into a problem that "page-up" does not work when we
reach the end of file:
https://sourceware.org/bugzilla/attachment.cgi?id=12201
The following fixes that, but breaks elsewhere (see 3):

@@ -232,14 +232,14 @@ tui_find_disassembly_address ...
          /* Disassemble forward a few lines and see ...
          next_addr = tui_disassemble (gdbarch, asm_lines, ...
          last_addr = asm_lines.back ().addr;
-         if (last_addr > pc && msymbol.minsym != nullptr
+         if (last_addr >= pc && msymbol.minsym != nullptr
              && asm_lines.size () >= max_lines)
            {
              /* This will do if we can't find anything... */
              possible_new_low.found = true;
              possible_new_low.new_low = new_low;
            }
-       } while (last_addr > pc && msymbol.minsym != nullptr);
+       } while (last_addr >= pc && msymbol.minsym != nullptr);

3. With the changes from above, "arrow up" stops working near
beginning of the file:
https://sourceware.org/bugzilla/attachment.cgi?id=12202

I can summarise what I have learned in the following table:

,-------------.----------.----------------------------------.
| usecase     | action   | condition that works             |
|-------------+----------+----------------------------------|
| very bottom | page-up  | as long as last_addr >= pc ...,  |
|             |          | go higher (try reducing new_low) |
|-------------+----------+----------------------------------|
| second line | arrow up | as long as last_addr > pc ...,   |
|             |          | go higher (try reducing new_low) |
`-------------^----------^----------------------------------'

This table portrays contradictory conditions for corner case
scenarios. I believe the real solution in the end should be
much simpler and ideally as such that it covers the corner
cases naturally. I will try to cook something up in my free
time.

I suggest that this patch series should be in along with
the changes proposed at point 1. So we only end-up with
none-working page-up at the end of assembly output.


Cheers,
Shahab


[1]
that x86_64 elf program is archived here:
https://sourceware.org/bugzilla/attachment.cgi?id=12199