Re: GDB 12.0.90 available for testing

[Joel Brobecker <brobecker@adacore.com>]

Hi everyone,

> Sorry for the slow reply, I've been off for the last week, so I'm only
> just catching up with my GDB emails.
> 
> How does the patch below look?  This moves the setbuf call into a new
> function, and includes the comment from gdb_readline_no_editing_callback.

Given the feedback received on this patch, and the fact that it was
tagged as important for the GDB 12 release, I performed one last round
of testing on Windows (using AdaCore's testsuite), and pushed Andrew's
patch below to both master and gdb-12-branch.

The one change I made was to the commit message, to add a reference
to the PR created by Eli.

Thanks everyone who contributed to finding, analyzing, fixing and
reviewing!

> commit 3ee791edccae840e11b9ebe70e5547dfbec04e00
> Author: Andrew Burgess <aburgess@redhat.com>
> Date:   Tue Apr 19 17:00:04 2022 +0100
> 
>     gdb: move setbuf calls out of gdb_readline_no_editing_callback
>     
>     After this commit:
>     
>       commit d08cbc5d3203118da5583296e49273cf82378042
>       Date:   Wed Dec 22 12:57:44 2021 +0000
>     
>           gdb: unbuffer all input streams when not using readline
>     
>     Issues were reported with some MS-Windows hosts, see the thread
>     starting here:
>     
>       https://sourceware.org/pipermail/gdb-patches/2022-March/187004.html
>     
>     The problem seems to be that calling setbuf on terminal file handles
>     is not always acceptable, see this mail for more details:
>     
>       https://sourceware.org/pipermail/gdb-patches/2022-April/187310.html
>     
>     This commit does two things, first moving the setbuf calls out of
>     gdb_readline_no_editing_callback so that we don't end up calling
>     setbuf so often.
>     
>     Then, for MS-Windows hosts, we don't call setbuf for terminals, this
>     appears to resolve the issues that have been reported.
> 
> diff --git a/gdb/event-top.c b/gdb/event-top.c
> index b628f0397cb..a55338d78a2 100644
> --- a/gdb/event-top.c
> +++ b/gdb/event-top.c
> @@ -818,19 +818,6 @@ gdb_readline_no_editing_callback (gdb_client_data client_data)
>    FILE *stream = ui->instream != nullptr ? ui->instream : ui->stdin_stream;
>    gdb_assert (stream != nullptr);
>  
> -  /* Unbuffer the input stream, so that, later on, the calls to fgetc
> -     fetch only one char at the time from the stream.  The fgetc's will
> -     get up to the first newline, but there may be more chars in the
> -     stream after '\n'.  If we buffer the input and fgetc drains the
> -     stream, getting stuff beyond the newline as well, a select, done
> -     afterwards will not trigger.
> -
> -     This unbuffering was, at one point, not applied if the input stream
> -     was a tty, however, the buffering can cause problems, even for a tty,
> -     in some cases.  Please ensure that any changes in this area run the MI
> -     tests with the FORCE_SEPARATE_MI_TTY=1 flag being passed.  */
> -  setbuf (stream, NULL);
> -
>    /* We still need the while loop here, even though it would seem
>       obvious to invoke gdb_readline_no_editing_callback at every
>       character entered.  If not using the readline library, the
> diff --git a/gdb/top.c b/gdb/top.c
> index 1cfffbeee7e..263a2ce8f43 100644
> --- a/gdb/top.c
> +++ b/gdb/top.c
> @@ -257,6 +257,41 @@ void (*deprecated_context_hook) (int id);
>  /* The highest UI number ever assigned.  */
>  static int highest_ui_num;
>  
> +/* Unbuffer STREAM.  This is a wrapper around setbuf(STREAM, nullptr)
> +   which applies some special rules for MS-Windows hosts.  */
> +
> +static void
> +unbuffer_stream (FILE *stream)
> +{
> +  /* Unbuffer the input stream so that in gdb_readline_no_editing_callback,
> +     the calls to fgetc fetch only one char at the time from STREAM.
> +
> +     This is important because gdb_readline_no_editing_callback will read
> +     from STREAM up to the first '\n' character, after this GDB returns to
> +     the event loop and relies on a select on STREAM indicating that more
> +     input is pending.
> +
> +     If STREAM is buffered then the fgetc calls may have moved all the
> +     pending input from the kernel into a local buffer, after which the
> +     select will not indicate that more input is pending, and input after
> +     the first '\n' will not be processed immediately.
> +
> +     Please ensure that any changes in this area run the MI tests with the
> +     FORCE_SEPARATE_MI_TTY=1 flag being passed.  */
> +
> +#ifdef __MINGW32__
> +  /* With MS-Windows runtime, making stdin unbuffered when it's
> +     connected to the terminal causes it to misbehave.  */
> +  if (!ISATTY (stream))
> +    setbuf (stream, nullptr);
> +#else
> +  /* On GNU/Linux the issues described above can impact GDB even when
> +     dealing with input from a terminal.  For now we unbuffer the input
> +     stream for everyone except MS-Windows.  */
> +  setbuf (stream, nullptr);
> +#endif
> +}
> +
>  /* See top.h.  */
>  
>  ui::ui (FILE *instream_, FILE *outstream_, FILE *errstream_)
> @@ -283,6 +318,8 @@ ui::ui (FILE *instream_, FILE *outstream_, FILE *errstream_)
>  {
>    buffer_init (&line_buffer);
>  
> +  unbuffer_stream (instream_);
> +
>    if (ui_list == NULL)
>      ui_list = this;
>    else
> @@ -412,6 +449,8 @@ read_command_file (FILE *stream)
>  {
>    struct ui *ui = current_ui;
>  
> +  unbuffer_stream (stream);
> +
>    scoped_restore save_instream
>      = make_scoped_restore (&ui->instream, stream);
>  
> 

-- 
Joel