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

Re: Please reconsider closure of # 457151 -- it affects gfortran transition

Colin Tuckley wrote:
> Kumar Appaiah wrote:
>> Actually, Colin has had a look at the package, and barring some finer
>> points, he said the package seems all right, except that we agreed
>> that we needed comments from someone (possibly Camm) before going
>> ahead with the upload.
>> Should anyone want to consider uploading it to experimental, please do
>> note the reason why it was rejected the last time; it was because the
>> blas3-paper.ps didn't have the source (it is a generated PostScript
>> file). Please REMOVE it before uploading it. Also, the version should
>> be 1.2-0.2, I guess.
> I'm planning on taking another look at this today and uploading to
> Experimental unless I find something to prevent it.
> Camm, any comments?
>> I was also of this view till some time back, though, of late, I have
>> started thinking that it is the package build system's responsibility
>> to build with minimal stuff even if `carrots' of extraneous
>> Build-Dependencies hang out around them. Of course, I would also be
>> very happy with the old system, but anyway... :-)
> In my view the best place to define an order for satisfying dependencies is
> in the "Depends:" statement. I think it would be best if the dpkg change was
> reverted ASAP and then we can have a discussion about the best way to move
> forward without having a lot of packages which FTBFS. Especially as this
> brokenness is currently impacting about 200 packages (via build dependencies
> on things like cernlib and refblas3) for armel.

The brokenness is in the packages that FTBFS, it was never guaranteed to
have dependencies installed in any particular order and it really
shouldn't be. The best thing to find bugs would probably to have it
randomly ordered or to test with all orderings, though that's a QA issue :-)

There is a bug in sbuild though which only looks at the first
alternative dependency it gets from dpkg-dev without checking if one of
the alternatives is already installed...



Reply to: