apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Erik Huelsmann <ehu...@gmail.com>
Subject Re: [PATCH] Don't define POSIX types {p,g,u}id_t
Date Tue, 26 Apr 2005 18:27:00 GMT
On 4/26/05, Cliff Woolley <jwoolley@virginia.edu> wrote:
> On Tue, 26 Apr 2005, Erik Huelsmann wrote:
> 
> > But do you consider breaking programs which depend on bugs in  the APR
> > implementation a bad thing?
> 
> Yes.  We are a library.  We *have to* maintain backward compatibility as
> outlined by http://apr.apache.org/versioning.html.  That is the guarantee
> we have made to those using our library.  Even if it means maintaining
> brokenness we accidentally included.

Then what's the use of doing any minor or patch releases? Programs
will come to depend on any bugs coded into 1.0.0 too.... They'll
compile, but if you fix the bug in 1.0.1, they'll stop working
correctly. Hardly any better than not compiling.

> If a program compiled on 1.0.0, it must also compile on 1.0.1.  Period.

I'm sorry to take my annoyance out on you, since it's not (only) you
who caused this (I'm getting an equally cooperative reaction with the
ruby team: "It's APR's fault: they should have used #define instead of
typedef.") But if you put it that way, there's hardly any reason to do
anything but major releases if you take the rules this litteral.

bye,


Erik (who starts to regret he ever tried to work this out).

Mime
View raw message