[dmd-internals] default dmd target == debdmd?

[Don Clugston]

On 27 August 2010 22:34, Steve Schveighoffer <schveiguy at yahoo.com> wrote:
>
>
>
>
> ----- Original Message ----
>> From: Don Clugston <dclugston at googlemail.com>
>> To: Discuss the internals of DMD <dmd-internals at puremagic.com>
>> Sent: Fri, August 27, 2010 4:27:13 PM
>> Subject: Re: [dmd-internals] default dmd target == debdmd?
>>
>> On 27 August 2010 17:28, Steve Schveighoffer <schveiguy at yahoo.com>  wrote:
>> >
>> >
>> >
>> >
>> > ----- Original Message  ----
>> >> From: Don Clugston <dclugston at googlemail.com>
>> >>  To: Discuss the internals of DMD <dmd-internals at puremagic.com>
>> >>  Sent: Fri, August 27, 2010 11:21:20 AM
>> >> Subject: Re: [dmd-internals]  default dmd target == debdmd?
>> >>
>> >> On 27 August 2010 17:15,  Steve Schveighoffer <schveiguy at yahoo.com> wrote:
>> >>  > I  was in the process of doing some windows phobos development, so I
>>decided
>> >>to
>> >> > start using the latest compiler from  trunk.  But when I started
>  building
>> >> > druntime and phobos, I was  getting these  printouts:
>> >> >
>> >> > Max # of fixups =  0
>> >> > Max # of fixups =  1
>> >> > Max # of fixups =  4
>> >> > Max # of fixups = 1
>> >> >  ...
>> >>  >
>> >> > I thought WTF?  Well, it turns out the default build of  dmd  if you just
>>do
>> >>make
>> >> > -f win32.mak is debdmd, not  dmd.
>> >> >
>> >> > Is  there a good reason for this?  If  not, can we change the default
>>target
>> >>to
>> >> >  dmd?
>> >>
>> >> It's so you can debug the compiler.Pretty much the  only  time you want
>> >> the release version is when making an official   release.
>> >
>> > Or when you want to utilize unreleased  fixes...
>>
>> I do not think we should encourage that. Nobody should really be  doing
>> that unless they are helping debug/test the compiler. If the fixes  are
>> that important, we should be doing a release.
>>
>> Note that even when  developing Phobos, you shouldn't ordinarily be
>> checking in code which depends  on the latest compiler bugfix. Every
>> time that has happened so far, a  compiler beta release has been made.
>> It's not reasonable to expect everyone  working on or testing Phobos,
>> to build the compiler  themselves.
>
> I understand this point of view, and I agree.  I won't use the trunk version in
> the future.  I still don't think the default should be debug mode, it's very
> confusing.  I can envision people building a released version of the source
> (i.e. from a downloaded tarball) and expecting it to be a release version
> without having to specify different arguments.  How much will this happen?  Not
> sure, but why provide the source if you don't want people to compile it...
>
> -Steve

It's often happened that someone hits a compiler segfault. By
compiling a debug version, and running:
windbg dmd foo.d
they can find which line it's segfaulting on. This has been quite
useful in the past; even people with absolutely no knowledge of C++
have been able to provide helpful info about bugs.
Hopefully, though, most of the segfaults are gone, now <g>.