1 |
On Sun, 17 Oct 2004, Tavis Ormandy wrote: |
2 |
|
3 |
> I maintain the at&t ksh package, app-shells/ksh, which is currently |
4 |
> failing to build for some people and not for others. |
5 |
> |
6 |
> The package works fine for me, and I can't reproduce any of the problems |
7 |
> other people are reporting, unfortunately the non-standard build process |
8 |
> is very complex and difficult to follow. |
9 |
> |
10 |
> I would appreciate any help in tracking this bug down, if anyone on this |
11 |
> list can spare a few minutes to try and build ksh and email me the the |
12 |
> make.out if it fails (preferably gzipped!) and the output of emerge info, |
13 |
> or if anyone can provide any insight on why this might be happening, I |
14 |
> would really appreciate it. |
15 |
|
16 |
This problem would be a *lot* easier to debug if someone wasn't passing |
17 |
-k to make. Due to this flag being used, the compile does not fail when |
18 |
the error happens, but rather when make runs out of other things it can |
19 |
work on. The chmod in question happens to be the last command the |
20 |
makefile can attempt if something goes wrong with building ksh, almost |
21 |
regardless of what goes wrong. |
22 |
|
23 |
I personally believe that, while make -k may be useful when trying to |
24 |
develop a project that requires components that are other team members' |
25 |
responsibility, but they have not yet finished them, it does not have a |
26 |
place in production releases. Using make -k in a released product is, |
27 |
to me, an admission that the person who released the product doesn't |
28 |
know about everything that goes on in the compile/install process. |
29 |
IMHO, if one is releasing a product, one should understand what it does, |
30 |
both while it's being built, and while it's running. |
31 |
|
32 |
Ed |
33 |
|
34 |
-- |
35 |
gentoo-dev@g.o mailing list |