Re: makeicecat fails for a miscalculation of the version of rename

[bill-auger]

i will try to explain - this issue is complicated on both ends


On Sun, 22 Jan 2023 15:30:44 +0100 chippy wrote:
> Arch maintains Icecat

to be clear, arch does not maintain an icecat package - that is
the AUR - everything on the AUR is maintained by arch users


On Sun, 22 Jan 2023 15:30:44 +0100 chippy wrote:
> This works on Debian but unfortunately not on a live instance of Arch
> (with perl-rename installed with pacman) where --nofullpath is an
> unrecognized option.

yes, that is known - it is the only reason why the check was
added; and why the error message specifies to use Parabola's
'perl-file-rename' package

the suggestion is possible; but i would not bother to support
the arch variant of perl-rename - it is non-standard; and only
arch has it - upstreams generally should not try so hard to
support specific opinionated downstreams - likewise, it would
have been cleaner for the AUR packager, to simply require
Parabola's 'perl-file-rename' package, rather than patching
the upstream build script to support arch

i could have implemented what you are suggesting originally,
but chose not to - the current implementation is cleaner - the
cost is but a minor nuisance, which only affects those who
package icecat for arch-like distros - AFAIK, that is exactly
two people

instead, i wrote to the arch packager, requesting to change the
non-standard and unmaintained variant to the standard maintained
variant - if that happens, parabola could drop the extra
package, and gnuzilla could safely delete the check for the
non-standard variant; because probably, no one would ever be
using it - ie: generally, a solution which allows you to delete
code, is the better than one that entails adding new code

lastly, there is no reason why any downstrem should ever need to
run `makeicecat` - up until a few years ago, gnuzilla always
published versioned source-balls for downstreams (AFAIK, GNU
requires all GNU projects to do so) - ive assumed all along,
that the habit of maintaining the project only as a VCS, was
temporary - that is another reason why it made little sense to
support arch users - if gnuzilla was still publishing
source-balls, no one would have ever noticed this bug; because
that code which checks for arch's incompatible `rename`, would
never have been written