Re: cygwin gcc 3.4 and cygwin

[Brian Ford <ford at vss dot fsi dot com>]

On Thu, 13 Mar 2003, Ronald Landheer-Cieslak wrote:

> On Wed, 12 Mar 2003, Brian Ford wrote:
>
> > Max Bowsher wrote:
> > >Brian Ford wrote:
> > >> I thought I had a legitimate concern and question, not one that
> > >> deserved "just" a sarcastic response.
> > >Yes, it was sarcastic, but don't take it personally. Chris is *busy* and
> > >this is quite a minor issue.
> > We are all *busy*.  Well, ABI breakage is not normally considered a minor
> > issue.
> Mr Faylor is the head of the Cygwin project. He might not be the
> papa-poule kinda guy and sweet-talk you to bed, but he does know what he
> is doing, especially when it comes to his work on Cygwin.
>
I know he is the head of the Cygwin project.  I didn't ask for sweet talk.
I simply stated a fact, asked a question, and advised caution.

Vanilla gcc 3.2.x is NOT ABI compatible with Cygwin's pre 3.2 gcc, fact.
I wanted to know why this was allowed to persist given the simple patch
required to fix it, question.  I had seen Mr. Faylor make statements
before to the effect of: doubles in structures are not that common so who
cares, so I advised caution in his use of that compiler for official
Cygwin released software.

> If mr Faylor tells you not to worry because Cygwin is safe, that means
> that you shouldn't worry because Cygwin is *safe*.
>
I have confidence that Mr. Faylor is a very competent individual.  We all
overlook things from time to time though, and his previous comments about
the issue in question were also sarcastic and not informative.  A single
declarative sentence would have addressed all my concerns, instead of the
consistent sarcasim I received.

> If mr Faylor tells you that he builds Cygwin with OOTB gcc on a regular
> basis (and he did so in response to me saying the gcc OOTB is not a
> working compiler for the Cygwin environment) that means that he's right
> and I was wrong. Even if it means that whatever he compiles with the OOTB
> gcc compiler is not compatible with whatever he compiles with the Cygwin
> edition of gcc - mr Faylor knows what he's doing, and he has known what
> he's doing for quite a while now.
>
He does know what he is doing, but others may misinterpret his statements
given that they do not have his global understanding of the issues
involved.  I was trying to prevent that.

Other people do release official Cygwin software packages.  The whole
project is a volunteer effort.

> > I know from following the list that I am just supposed to accept Chris
> > this way.  But honestly, I'd rather receive no reply than one with just
> > sarcasim.  One with sarcasim and real content would be fine.  It takes
> > nearly the same amount of time to do both.
> Like I (and many others) said: he's not one to sweet-talk you to bed - you
> can ask your mom & dad to do that.
>
This is uncalled for.  Again, I did not ask for sweet talk, just a simple
clarification.

> >>> It would be easy to accendentally release things for Cygwin that are
> >>> ABI incompatible with Cygwin's gcc.
> >> structs containing doubles aren't a hugely common feature. Besides, I
> >> think Chris knows what he is doing.
> > Many libraries might be affected, especially ones that deal with
> > images, etc.
> Have there been reports of bugs related to this ABI breakage lately? hmm?
>
Just mine :)

> > BTW, this may not be the only issue, it is just the only one I have
> > bumped into so far.  I wish I could feel comfortable that it is the only
> > one, but with responses like this, I don't feel warm and fuzzy about
> > anything Cygwin.
> If you find something broken, take a look at
> http://www.cygwin.com/bugs.html and report it.
> If you can fix it, feel free to do so.
>
I do.  Both.

> [snip legwork for gcc]
>
> Frankly, as long as mr Faylor doesn't say he's had it with Cygwin and
> wants to take more time to do, say, French cuisine (do you cook, Chris?
> It's fun!) I'd say Cygwin is in good hands, and wouldn't worry too much
> about ABI incompatibility and other such nonesense.
>
We may have significant in-house tools depending on Cygwin.  Our
developers are also competent individuals, and my company is not
opposed to donating large improvements/bug fixes when they are mutually
beneficial.  Furthermore, we may end up becoming a Redhat customer for
Cygwin.  Thus, we do care about the quality of this volunteer effort and
are prepared to back that up with our own contributed analysis and
development.

It would be a shame to loose quality contributors to Cygwin's volunteer
effort just because current project leaders leave bad impressions.

-- 
Brian Ford
Senior Realtime Software Engineer
VITAL - Visual Simulation Systems
FlightSafety International
Phone: 314-551-8460
Fax:   314-551-8444




--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/