Re: [PATCH 0/9] Add UBSan to the build

[Pedro Alves <palves@redhat.com>]

On 08/30/2018 07:55 PM, Tom Tromey wrote:
>>>>>> "Pedro" == Pedro Alves <palves@redhat.com> writes:
> 
> Pedro> Might be noticeable in "make check" time.
> Pedro> If you're doing performance analysis, say, running "make check-perf", or running
> Pedro> some use case under "perf", sounds like you'll need to be sure to
> Pedro> disable UBSan.  I also mildly worry about random people comparing the
> Pedro> performance of master GDB or some ftp snapshot against previous GDB versions
> Pedro> or against other debuggers and being mislead.  If that slowdown is true, I think we
> Pedro> should at least document it somewhere more prominently, as I think that so
> Pedro> far release vs development mode didn't have much of an impact?
> 
> Anyone doing performance analysis really must compile with -O2 and
> disable all the checking.  It really does matter, IIRC the sleb/uleb
> decoders really suffer without optimization.

I think optimization/-O2 is much more obvious.  Enabling/disabling
the checking so far hasn't been an issue so far, I don't think.

> 
> Pedro> Kind of like how GCC documents --enable-checking, at
> Pedro> <https://gcc.gnu.org/install/configure.html>:
> 
> Pedro> "When you specify this option, the compiler is built to perform internal consistency
> Pedro> checks of the requested complexity. This does not change the generated code, but adds
> Pedro> error checking within the compiler. This will slow down the compiler and may only work
> Pedro> properly if you are building the compiler with GCC. This is ‘yes,extra’ by default when
> Pedro> building from SVN or snapshots, but ‘release’ for releases."
> 
> Do we have a similar doc somewhere?  I could update it.

There's a "configure options" node in the manual.

And also a "configure' options" section in gdb/README too.

> Also I think this series should update NEWS for the new configure option.

/me nods.

Thanks,
Pedro Alves