"Dr. Thomas Orgis" <thomas.orgis%uni-hamburg.de@localhost> writes:
> Because of that, this patch is needed (you see how old it already
> is:-/):
>
> --- pkgsrc-2015Q4/security/mit-krb5/builtin.mk 2014-02-28 13:17:20.000000000 +0100
> +++ pkgsrc/security/mit-krb5/builtin.mk 2015-11-04 15:55:49.281124121 +0100
> @@ -10,9 +10,12 @@
> BUILTIN_FIND_HEADERS.H_MIT_KRB5= krb5/krb5.h
> .elif !empty(MACHINE_PLATFORM:MSunOS-*-*)
> BUILTIN_FIND_HEADERS.H_MIT_KRB5= kerberosv5/krb5.h
> +.elif !empty(MACHINE_PLATFORM:MLinux-*)
> +# Assuming mit-krb5 >= 1.5 on GNU/Linux.
> +BUILTIN_FIND_HEADERS.H_MIT_KRB5= krb5/krb5.h
> .else
> BUILTIN_FIND_HEADERS.H_MIT_KRB5= krb5.h
> .endif
> -BUILTIN_FIND_GREP.H_MIT_KRB5= Massachusetts Institute of Technology
> +BUILTIN_FIND_GREP.H_MIT_KRB5= Massachusetts
> BUILTIN_FIND_FILES_VAR:= SH_KRB5_CONFIG
> BUILTIN_FIND_FILES.SH_KRB5_CONFIG= /usr/bin/krb5-config
That looks entirely reasonable to me. (These ought to be committed
seaprately anyway.)
> --- a/security/mit-krb5/builtin.mk
> +++ b/security/mit-krb5/builtin.mk
> @@ -88,6 +88,10 @@ ALL_ENV+= KRB5_CONFIG=${KRB5_CONFIG:Q}
> BUILDLINK_CPPFLAGS.mit-krb5!= ${SH_KRB5_CONFIG} --cflags
> BUILDLINK_LDFLAGS.mit-krb5!= ${SH_KRB5_CONFIG} --libs
>
> +. if !empty(SH_KRB5_CONFIG:M/usr/lib/mit/*)
> +BUILDLINK_PREFIX.mit-krb5= /usr/lib/mit
> +. endif
> +
> #
> # The SunOS builtin krb5-config does not support all of the arguments that the
> # MIT version does so we install a fake script which strips them out.
> I guess this has been communicated on IRC some time. I do not see that
> _I_ need this change (no /usr/lib/mit/), but maybe it I got it for a
> reason? Jonathan?
I would prefer that we not commit patches unless someone can explain why
they are necessary and why they are correct. A prefix of /usr/lib/mit
seems like an odd thing to do (vs /usr/mit being a prefix and
/usr/mit/{include,bin,lib} being inside it. However I see there is a
reference to /usr/lib/mit/bin already in builtin.mk!
If prefix is wanted, why doesn't the code use "krb5-config -prefix" once
found? Perhaps that's because some krb5-config implementations are
deficient per the comments.
I don't see how BUILDLINK_PREFIX.mit-krb5 is getting set in the normal
case. Probably that's implicit in some other mk fragment...
So this minute I am opposed to committing this, but I am very willing to
be talked out of that position by additional information. Right now we
don't seem to know when /usr/mit/lib might appear and if that is still
relevant.
Attachment:
signature.asc
Description: PGP signature