On 2007-1-27 18:30 UTC, Keith Marshall wrote:
> On Friday 26 January 2007 01:05, Greg Chicares wrote:
>> On 2007-1-26 0:37 UTC, Earnie Boyd wrote:
>> > We already have Snapshots. The packages currently in Proposed are for
>> > the most extent Unmaintained.
>> Okay, I agree: of eight "Proposed" packages, only one is dated
>> within the last twelve months, and half are from 2003.
[snip discussion of two of the more recent packages]
> On the remaining six, I don't wish to comment; perhaps the original
> contributers would care to do so, if they are reading this.
As for mpatrol:
I guess I'm the contributor in a sense, and I'm willing to assume
responsibility for maintaining it somehow--but I wonder whether a
package is the right way to do that. It seems that this library has
to be custom-built for whatever combination of gcc and binutils is
installed, because it replaces the C and C++ memory-allocation
libraries and depends on libiberty and libbfd. There have been
conflicts between gcc and binutils with respect to those libraries:
which must be managed carefully for mpatrol to work.
The package has value as source, but presumably works for only one
2003 combination of gcc and binutils, and otherwise must be rebuilt:
| OK, I finally figured out mpatrol.
| The precompiled version on the mingw site is "broken"
| grab the source at the mingw site which they already patched
| (better too since it generates dlls).
I'm continuing to develop my own procedures for navigating these
difficulties so that my coworkers can use mpatrol. We've mandated
its use for our regression tests, because it really does find some
nasty problems with little effort; so we're making sure it keeps
working with MinGW tools.
What's the best way to offer this work to the community, in light
of the dependency on both gcc and binutils versions?
- a versioned package to accompany each MinGW-installer release?
- autotoolize the hand-written makefiles we now use?