On Thu 21 Jul 2022 at 11:47:58 (-0500), Igor Korot wrote:
> On Thu, Jul 21, 2022 at 10:31 AM David Wright <deb...@lionunicorn.co.uk>
> wrote:
> > On Thu 21 Jul 2022 at 10:15:43 (-0400), Greg Wooledge wrote:
> > > On Thu, Jul 21, 2022 at 09:06:53AM -0500, David Wright wrote:
> > > > I thought that was what the attached was (actually for Ubuntu AIUI).
> > > > As I originally wrote, "As if by magic, […] someone else supplies a
> > > > copy."
> > > >
> > >
> > > > #! /bin/sh
> > > >
> > > > # This shell script saves various pieces of information about the
> > > > # installed version of unixODBC. Packages that interface to
> > > > # unixODBC can use it to configure their build.
> > > > # This file replaces the standard odbc_config, which is not
> > > > # relocatable
> > > > #
> > > > # Author: Alberto Di Meglio <alberto.di.meg...@cern.ch>
> > > > # Public domain
> > > >
> > > > me=`basename $0`
> > > > mydir=`dirname $0`
> > > > mydir=${mydir%/bin}
> > > >
> > > > # stored configuration values
> > > > val_prefix="$mydir"
> > > > val_bindir="$mydir/bin"
> > > > val_includedir="$mydir/include"
> > > > val_libdir="$mydir/lib"
> > > > val_libs="-L$mydir/lib -lodbc"
> > > > val_version='2.2.11'
> > >
> > > This is not going to give the correct linker arguments. It's going to
> > > spit out a -L option which is totally unneeded, and worse, the *content*
> > > of that -L option is going to depend on where the operating system thinks
> > > the script has been "installed".
> > >
> > > If the script is "installed" in /usr/local/bin/odbc_config, it's going
> > > to spit out -L/usr/local/lib -lodbc.
> >
> > Yes, there's a hint in the file: "This file replaces the standard
> > odbc_config, which is not relocatable". It may be that this script
> > is sensitive to where it is placed, even though that was not the
> > intention. That alone might cause it, or something else, to fail.
> >
> > It's also quite possible that the OP copied it into …/pkg-config,
> > ran that, and reported "it fails." Who knows?, hence my complaint:
> > "whether you have taken any actions as a result of reading the
> > URLs that have been mentioned".
> >
> > > As the script itself says,
> > >
> > > > # This file replaces the standard odbc_config
> > >
> > > So, why not use the "standard odbc_config", whatever that is?
> >
> > We're told that it's not part of the Debian distribution.
> >
> > > Again, this really needs to be taken up with the upstream maintainers of
> > > the library, and with the Debian maintainer(s) of the Debian packages of
> > > the library.
> >
> > I think it kind of was, in the Debian BTS that I referenced and
> > advised the OP to read carefully. I can say no more because,
> > unlike you, my experience of compiling C runs to not much more
> > than including a couple of libraries and producing an a.out.
> > (Two decades ago, I compiled kernels, but kernel-package made
> > that trivial apart from deciding which items to include.)
>
> So, I got a reply from Nick Gorham (maintainer of unixODBC).
>
> In the E-mail he said:
>
> [quote]
> I think pkg-config is made during the build time to match the distribution"
> [/quote]
I don't know what the significance of such a one-line quote is.
I can only note that there are two Debian packages that already
include /usr/bin/pkg-config programs.
> Which means that I will be needing something in the configure
> to check if odbc_config exists, use it if it is and if not - use
> pkg-config.
>
> It still feels wrong however, that the distro maintainers dictate
> to upstream packages how to do their business.
I wasn't aware that they did. Distributions have policies, and AFAIK,
maintainers use patches to modify the upstream sources to fit in with
those policies.
> 1. vWhere can I file the bug about missing odbc_config?
Presumably in the Debian BTS, under the package that you're reporting
it's missing from.
> 2. Is there a ML/forum/whatever to discuss the issues like this?
> This forum is for users of the OS, not people who wants to develop
> software on the OS.
The unixodbc homepage mentions a mailing list that has been migrated
and fixed, but the entry's timestamp (2008) doesn't inspire confidence
that it's still active. Did the Debian maintainer not mention it?
Cheers,
David.
