Hi Phil,

The short answer is that I already deal with enough "smart" compilation tools 
and I'm afraid that I will be finding corner case after corner case in this 
one. The mechanism needs to be as dumb as possible.

There is a commercial software package for seismic data acquisition and 
processing that I support at my site, called Antelope [1]. This package lives 
in it's own little tree under /opt/antelope, largely independent of system 
software.

The major exception to the self-contained nature of Antelope is the contributed 
code library[2], which has dependencies on a lot of open source projects - 
things like ImageMagick, RRDTool, Python, etc. My site makes heavy use of 
"antelope_contrib" because almost all of my users are Power Users and many are 
authors of items in antelope_contrib.

Maintaining the products generated by compiling the antelope_contrib tree 
unfortunately is where most of my time seems to go because of the third-party 
and open source dependencies. There are a number of eccentricities with the 
build process for the antelope_contrib code, and the end result is that I have 
to battle with several "smart" tools that set and tweak various environment 
variables.

I had a bit of a masochistic laugh when I read your response [3] to me on 
maintainers that used a command called "compile" to perform the auto-switch - 
that is EXACTLY the name of the goofy macro/wrapper that this commercial 
package uses to abstract between SPRO and GCC on Solaris. I think the 
AntelopeMake Makefile actually sets "CC" to "compile" during the build phase.

I'd really prefer to set and forget using the alternatives mechanism, rather 
than have a tool try to guess my intentions through three or more layers of 
abstraction. Using something like the alternatives mechanism, I can force my 
build process to look at the "right" auto-configure tool for a package like 
Python or Ruby or RRDtool.

[1] http://brtt.com
[2] http://github.com/antelopeusersgroup/antelope_contrib
[3] http://lists.opencsw.org/pipermail/maintainers/2011-March/014365.html

On Mar 21, 2011, at 1:36 PM, Philip Brown wrote:

> Hello there,
> I'm keeping an eye on the polling thing :)  and I'm curious, why you chose
> "site wide, fixed configuration"
> rather than having the tool auto-switch based on value of CC?
> 
> I'm having difficulty understanding why, in cases where the tool can
> easily suport both, and it will default to (something reasonable) if
> CC is not set... why you would not prefer it to be auto-switching?

_______________________________________________
maintainers mailing list
[email protected]
https://lists.opencsw.org/mailman/listinfo/maintainers
.:: This mailing list's archive is public. ::.

Reply via email to