[lfs-dev] LFS 7.10: 2.2. Host System Requirements: Perl version OK: module missing
lfs65 at cruziero.com
Tue Dec 13 13:11:22 PST 2016
> From: Bruce Dubbs <bruce.dubbs at gmail.com>
> Date: Tue, 13 Dec 2016 10:36:20 -0600
> Subject: Re: [lfs-dev] LFS 7.10: 2.2. Host System Requirements: Perl version
> OK: module missing
> akhiezer wrote:
> >> From: Bruce Dubbs <bruce.dubbs at gmail.com>
> >> Kevin Buckley wrote:
> >>> A bit further into the build, though, I was told that I didn't have the
> >>> Getopt::Std.pm
> >>> module.
> >>> Easy enough to fix and whilst it might be "overkill" for the book, which
> >>> probably assumes that folk have a build host that's not as stripped down
> >>> as a debootstrapped OS install, it struck me that might be worth adding
> >>> in an explicit check for that one Perl module ?
> >> The file, AFAICT is /usr/lib/perl5/*/Getopt/Std.pm.
> >> Why did you strip that out? This is essentially a library. We don't
> >> search for every library that we need and Getopt/Std.pm is part of a
> >> standard perl install.
> >> I am against adding this to version-check.sh because it is too much of an
> >> edge case, but I am open to other opinions.
> > Can be useful in helping automate host-os requirements: strace'ing a
> > build shows - at least in theory - all that is accessed of host-os;
> > although strace can o/c interfere with what it's tracing.
> I'm willing to do it if someone can show me a distro where a
> apt-get/yum/dnf etc install of perl does not install Getopt::Std.pm.
Just to be clear: the strace mention, was in case the include-all route
was opted for; and isn't a recommendation to go down the route.
> In the same vein, I am thinking about removing the library consistency
> check because that was a similar one-off problem where the solution tends
> to confuse more than it helps.
In light of the new instance, 'Getopt/Std.pm', pointing to the general
issue of how much to include in the version-check page: I'd agree
broadly that one doesn't want to treat libraries/&c (if at all) in the
same way as the other 'main' items in version-check.
If they are to be omitted, then maybe a small para/item on the page
noting the issue, could be useful; although it too perhaps could
'tend[s] to confuse more than it helps'. Maybe just say something
that includes s'thing like, roughly, "[...] most testing with full,
and not heavily-modified, installs of the main well-known distros;
and less tested with other distros or roll-your-own systems.
However, staying with the 'include-all' idea for a moment: might
version-check be extendable and generalised to a './configure ...'-style
test for what's on the host-os system -vs- what is known to be required.
More information about the lfs-dev