Subject: Re: buildink2 hides a wee bit too much -- it can no longer detect inconsistencies
To: None <email@example.com>
From: Greg A. Woods <firstname.lastname@example.org>
Date: 11/21/2002 11:39:56
[ On Thursday, November 21, 2002 at 00:32:20 (-0800), Johnny C. Lam wrote: ]
> Subject: Re: buildink2 hides a wee bit too much -- it can no longer detect inconsistencies
> The error you are discussing will occur only if the user "cvs updates"
> www/amaya without "cvs updating" www/libwww. The libwww buildlink2.mk
> file specifies that it requires libwww>=5.4.0, presumably because the
> shared libs in 5.4.0 are ABI-incompatible with shared libs in 5.3.6.
> This implies that all packages that need libwww, including www/amaya,
> will require libwww>=5.4.0.
> Given how interdependent the whole pkgsrc tree is, it's a bug to do
> selective updates to parts of the pkgsrc tree.
I this case it was a selective update, but that's far from the only way
to get inconsistencies like that from the head of a CVS working branch.
I think it's really important to be explicit about inter-package
> > Ultimately this is caused by buildink2 hiding too much of the
> > inter-package dependency declarations in unrelated directories (or at
> > least making it too easy for the developer to do such hiding).
> I disagree. I believe it is due to the mistaken assumption that you can
> selectively update parts of the pkgsrc tree. I also think this is the
> chance that you take by tracking pkgsrc-current because it's a _very_
> active branch and it's possible to "cvs update" between two related
So you would consign all developers (including third party developers)
to this uncertainty just because it saves one simple declaration?
Greg A. Woods
+1 416 218-0098; <email@example.com>; <firstname.lastname@example.org>
Planix, Inc. <email@example.com>; VE3TCP; Secrets of the Weird <firstname.lastname@example.org>