bug-lilypond
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Issue 1905 in lilypond: Add note to CG 9.3 'Compiling reg tests' to


From: David Kastrup
Subject: Re: Issue 1905 in lilypond: Add note to CG 9.3 'Compiling reg tests' to add --disable-optimising when using autogen.sh or configure
Date: Mon, 19 Sep 2011 18:27:59 +0200
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/24.0.50 (gnu/linux)

"address@hidden" <address@hidden> writes:

> On Sep 19, 2011, at 5:22 PM, Neil Puttock wrote:
>
>> On 19 September 2011 16:16, David Kastrup <address@hidden> wrote:
>> 
>>> The debug symbols should not affect code speed, and they are present in
>>> the unoptimized build, anyway, unless you are talking about something
>>> completely different from what I think you do.  I find it disturbing,
>>> however, that our default build uses NDEBUG to disable assertions.
>>> Under normal circumstances, assertions have negligible speed impact.  It
>>> makes little sense to restrict their use explicitly to non-optimized
>>> builds.  In particular since it makes it quite unlikely that one can
>>> catch Heisenbugs with assertions: they often go away with significant
>>> code changes, and switching optimization off most certainly _is_ a
>>> significant change.
>> 
>> I didn't say that the debug symbols affect code speed.
>> 
>> The unoptimized binary is much slower, at least on my system.
>> 
>> Mike is suggesting we stop shipping optimized binaries.
>> 
>
> I've put up a new patch that checks for cyclic dependencies in the
> regtests - does this seem like a better approach?

It does not really touch the issue of using NDEBUG on default builds.
For what it is worth: I think that Lilypond is the only project I
remember working with where NDEBUG (and consequently disabled
assertions) are the default.

This particular issue is concerned about regtests/--disable-optimising,
so I can't quite see how the cyclic dependency patch would be concerned
unless the check is only done with NDEBUG set, at considerable cost.

-- 
David Kastrup



reply via email to

[Prev in Thread] Current Thread [Next in Thread]