Re: Build-Depends question re: yacc, ctags, etc.
arto.astala@nokia.com writes:
> If you know, or suspect, that some might not do then make
> that knowledge explicit (at least by not listing all similar
> packages). All the packages providing *tags are probably
> pretty equivalent, but yacc, byacc, bison etc. have differences
> that mean that some grammars are not acceptable to all of those.
Hmmm, but if the upstream Makefile invokes it as ``yacc'', wouldn't
any of those yacc-providers suffice? If not, we've got a problem: If
I say ``Build-Depends: bison'' it only requires that bison be
*installed* on the build system, not that it be used when the Makefile
calls for yacc -- the builder's yacc alternative may point to byacc.
Should we be worried about this?
> Should we require in policy that sources have a text file
> explaining the dependencies. This would be a natural place
> for telling whether all of the 'equivalent' packages do or
> not.
For whom to read? I assume the build daemons or humans, or eventually
the users, would file a bug report if the specified Build-Depends
isn't complete or results in incorrect behavior. I suspect we don't
want to ask the builders to "figure it out for yourself" (though maybe
I'm wrong, because they've had to do that in the past).
We might do well to advise maintainers to keep notes about those
dependencies (in the changelog or in control with something like
``X-Build-Dependencies-Comments: ...'') for later reference, but I
think this is no different from other sensible maintainer
responsibilities.
Reply to: