Control sequences output when testing GDB

[Simon Marchi]

On 2020-03-17 9:51 a.m., Simon Marchi wrote:
> Hi,
> 
> I had to make a GDB fix related to Cygwin recently, so I tried to run some tests
> from the GDB testsuite on Cygwin.  This is my first foray into Cygwin.
> 
> The test I ran failed, because the GDB under test outputs what appears to be control
> sequences to clear the remaining characters of the line and go to the new line.  What's
> surprising is that these control characters only start to appear when we use the "run"
> command of GDB, which makes GDB spawn the process to debug.
> 
> The full test log is here:
> 
>   https://pastebin.com/raw/m6JYGRvs
> 
> But here are a few lines around where it starts getting interesting:
> 
>   (gdb) break main
>   Breakpoint 1 at 0x1004012e3: file /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/sizeof.c, line 101.
>   (gdb) run
>   Starting program: /home/smarchi/build/binutils-gdb/gdb/testsuite/outputs/gdb.base/sizeof/sizeof
>   [?25lType "apropos word" to search for commands related to "word".
>   (gdb) set height 0
>   (gdb) set width 0
>   (gdb) dir
> 
> The numbers match the number of characters left to the line, after what has already been
> output (for a total of 80).  According to [1], the 'X' is "Erase Character" and the 'C' is
> "Cursor Forward".
> 
> When the program starts, we also see one for "Text Cursor Enable Mode Hide".
> 
> These control sequences are a bit problematic, because all the text that the testsuite
> is supposed to match is not matched, and the tests fail.
> 
> It was suggested to me (by jturney on IRC) to use:
> 
>   export CYGWIN=disable_pcon
> 
> To disable to use the new new Windows Pseudo Console thing.  And indeed, it makes the output
> free of these escape sequences, and the test passes fine.
> 
> Anybody can explain what is happening, and what we should do about it?
> 
> Thanks,
> 
> Simon
> 
> [1] https://docs.microsoft.com/en-us/windows/console/console-virtual-terminal-sequences
> 

Oh well, I just saw this response from Ken earlier today (sorry about
that, I'm not subscribed to the list):

  https://sourceware.org/pipermail/cygwin/2020-March/244091.html

So I guess there's no explanation yet, just the workaround.

Simon