Bug#962401: netcdf-fortran: please make the build reproducible

[Sebastiaan Couwenberg]

On 6/8/20 12:49 AM, Chris Lamb wrote:
>>> You have removed some settings, but you left in FFLAGS which varies on
>>> the build path.
>>
>> This is ideally something that is fixed in dpkg-buildflags or gcc, since
>> the introduction of the prefix-map option it also raised
>> file-references-package-build-path lintian issue for many packages.
> 
> Thanks for applying the patch, but I believe you have the cause and
> effect reversed in your thinking about this system.
> 
> The entire point of dpkg-buildflags is to pass the -ffile-prefix-map
> and -fdebug-prefix-map arguments to GCC. There is nothing to fix as
> their entire purpose is to pass this varying path to GCC. The only
> "fix" possible would be to remove this mechanism entirely, defeating
> the goal of being able to do a reproducible builds in the first place.
> 
> In other words, if a package happens to trigger that Lintian tag, then
> that is (correctly) informing the maintainer that it is (incorrectly)
> recording this inconsequential build path in a binary artefact and
> is something for the maintainer to address.
> 
> Hope that clarifies things on your end, despite this being somewhat out
> of scope for this bug report.

Recording buildflags is a very common practice by upstreams, it's not
the responsibility of the package maintainer to filter out buildflags
that contain the buildpath, that's something added by the toolchain. Not
using the buildpath for the prefix-map seems like a better idea. Or
having a dh tool remove the flag in post-processing, instead of
expecting maintainers to add patches like this.

While the intentions of the prefix-map may be good, the (lintian) fall
out of the implementation makes it no so good in my opinion.

Kind Regards,

Bas

-- 
 GPG Key ID: 4096R/6750F10AE88D4AF1
Fingerprint: 8182 DE41 7056 408D 6146  50D1 6750 F10A E88D 4AF1