| 1 |
On Thu, 02 Mar 2006 11:35:34 -0600 Lance Albertson |
| 2 |
<ramereth@g.o> wrote: |
| 3 |
| QA shouldn't have to depend on the tools you use. |
| 4 |
|
| 5 |
Sure. However, the tree is far too large to check manually for many |
| 6 |
things. If we were to do the Sekrit Tool's IUSE check manually, for |
| 7 |
example, we'd still be in app-something, and would have missed many of |
| 8 |
the screwups. |
| 9 |
|
| 10 |
| The final say should |
| 11 |
| be the human interaction. If doing weird white spaces breaks the tool, |
| 12 |
| but really isn't a QA issue outside of neatness, it shouldn't be |
| 13 |
| waving red flags. |
| 14 |
|
| 15 |
The problem is, without fixing the syntax weirdness it's not possible |
| 16 |
to tell whether red flags should be being waved for something else. |
| 17 |
|
| 18 |
| Yes, its probably something that should be fixed, |
| 19 |
| but it shouldn't be a critical one just because the tool is broken |
| 20 |
| and can't handle the weirdness. |
| 21 |
|
| 22 |
That's the thing... Doing static analysis on bash code is ludicrously |
| 23 |
difficult. If you don't believe me, try writing a tool that will figure |
| 24 |
out all ebuilds that have a redundant src_compile. |
| 25 |
|
| 26 |
It's a heck of a lot easier to do if you assume that developers will |
| 27 |
use sane syntax. Where developers don't use sane syntax, the only way |
| 28 |
to deal with it is to check it by hand. We don't have enough developers |
| 29 |
to do that. |
| 30 |
|
| 31 |
-- |
| 32 |
Ciaran McCreesh : Gentoo Developer (Wearer of the shiny hat) |
| 33 |
Mail : ciaranm at gentoo.org |
| 34 |
Web : http://dev.gentoo.org/~ciaranm |
Archives