| 1 |
I've noticed in quite a number of ebuilds that they fail one or more steps |
| 2 |
of their compilation from time to time. |
| 3 |
As an example, I'll take the most recent libgd package. |
| 4 |
Part of the unpack stage does this: |
| 5 |
|
| 6 |
cd ${S} |
| 7 |
patch -p <${FILESDIR}/gd-2.0.1.patch |
| 8 |
|
| 9 |
Either or both of these could fail. |
| 10 |
In my opinion, either all ebuilds need to be treated as though they were |
| 11 |
bash scripts with '-e' set (I think I can make a case for this), which means |
| 12 |
that *all* failures are treated as fatal (and why wouldn't they be?), |
| 13 |
or that *every* line end in "|| die" |
| 14 |
|
| 15 |
Either it's OK for patch to fail or it isn't. |
| 16 |
Either it's OK for the directory to exist or it isn't. |
| 17 |
It's common practice to use || die with make and configure, |
| 18 |
why isn't it with other, just-as-critical parts of the unpack;build;install process? |
| 19 |
|
| 20 |
|
| 21 |
-- |
| 22 |
Pound for pound, the amoeba is the most vicious animal on earth. |
| 23 |
|
| 24 |
Jon Nelson <jnelson@×××××××.net> |
| 25 |
C and Python Code Gardener |
Archives