[prev in list] [next in list] [prev in thread] [next in thread] 

List:       crux
Subject:    Re: Errors during Compilation
From:       Markus Ackermann <maol () symlink ! ch>
Date:       2002-06-17 16:29:22
[Download RAW message or body]

On Mon 17.06.2002,  9:31:04 -0600, Andrew Green said in public:
>You make it sound like there would be 50 or more return codes.  Looking 
>at a few sample Pkgfiles from base and contrib, it's apparent that there 
>are only a very FEW steps that exist for virtually every package:
>
>cd targetdir
>./configure
>make
>make install

Right, maybe more than 80% of all packages install that way. But it's the
other 20% I'm concerned about. Of course you could always throw in No. 7
"unknown error", but then... oh well, you could.

>Everything else can be user-defined, with a simple explanation in the 
>Pkgfile comment section.   I don't find it to be necessarily cluttered. 
> I find it to be a small sacrifice if you want to increase the 
>robustness of pkgmk.  Personally I find it astonishing that it has no 
>error checking at all.

>Seems people are interested in having this functionality, otherwise it 
>wouldn't be aired on this list.  This very thing has been discussed at 
>various times in #crux as well.
[...]
>change, was adopted, and then removed....But offering suggestions to 
>fixing the brokenness of a flawed-design is treated as if it's the 
>dumbest idea ever.

I'm one of the people who'd really like to have such a feature.

>If you had a more compelling argument than "it looks ugly", perhaps I 
>wouldn't be so dismissive of your statements.

It's because Makefiles (or ant scripts or whatever) and their
language/syntax are specifically designed for this task. Why try to model an
incomplete version of this mechanism with shell commands?

Markus.

[prev in list] [next in list] [prev in thread] [next in thread]