On 03/09/2010 02:02 PM, Gary Benson wrote:
Andrew Haley wrote:
> On 03/09/2010 12:53 PM, Matthias Klose wrote:
>> On 05.03.2010 17:31, Alexander Kurtakov wrote:
>>> Time to bring GCJ support discussion back.
>>> We should find a way to not install java-1.5.0-gcj on every
>>> users computers. There are packages that do not work with it
>>> and it's plain wrong to install it as a dependency of this
>>> packages which are explicitly requiring java 1.6.0.
>>> Possible solutions:
>>> * Solution 1 - someone who cares for the gcj stack finds a way
>>> to not require java-1.5.0-gcj when there are gcj bits in the
>>> package so we do not force gcj installation on every user.
>
> That shouldn't be too difficult. I'm looking at jakarta-oro now: I
> removed the
>
> --- jakarta-oro.spec~ 2009-07-26 13:52:21.000000000 +0100
> +++ jakarta-oro.spec 2010-03-09 13:24:34.445414981 +0000
> @@ -59,8 +59,6 @@
>
> %if %{gcj_support}
> BuildRequires: java-gcj-compat-devel
> -Requires(post): java-gcj-compat
> -Requires(postun): java-gcj-compat
> %endif
>
> %description
>
> and it's fine. There's no need AFAICS for a Requires: on java-gcj-compat.
If it uses rebuild-gcj-db in its %post and %postun scripts then it
needs those dependencies.
Gary, I just *tried it* !
No it doesn't. The actual scripts are
%if %{gcj_support}
%post
if [ -x %{_bindir}/rebuild-gcj-db ]
then
%{_bindir}/rebuild-gcj-db
fi
%endif
%if %{gcj_support}
%postun
if [ -x %{_bindir}/rebuild-gcj-db ]
then
%{_bindir}/rebuild-gcj-db
fi
%endif
So, if gcj is installed the right thing happens and if gcj isn't installed
the right thing still happens.
Andrew.