| 1 |
On Tue, Oct 25, 2005 at 08:57:59PM +0100, Ciaran McCreesh wrote: |
| 2 |
> On Tue, 25 Oct 2005 14:51:26 -0500 Brian Harring <ferringb@g.o> |
| 3 |
> wrote: |
| 4 |
> | portage substitutes "" when metadata keys are unset during depends |
| 5 |
> | export. Seems kind of pointless requiring an empty var |
| 6 |
> | <snip> |
| 7 |
> | to do the check |
| 8 |
> | would require mangling ebuild.sh also, which I think is kind of daft. |
| 9 |
> |
| 10 |
> Hrm. KEYWORDS.missing, LICENSE.missing, DESCRIPTION.missing and |
| 11 |
> SLOT.missing are listed in 'man repoman'. |
| 12 |
|
| 13 |
Re-read the chunks of my original email above. |
| 14 |
|
| 15 |
The suggestion is requiring |
| 16 |
|
| 17 |
HOMEPAGE='' |
| 18 |
|
| 19 |
in ebuilds, which is daft due to the fact that ebuild.sh does not |
| 20 |
discern between an unset and empty var for exporting. All stable |
| 21 |
portage (python side) sees is empty, no special flagging of unset. |
| 22 |
|
| 23 |
The repoman checks I commented on, due to ebuild.sh not discerning |
| 24 |
between an unset env var and an empty one, are testing to see if it's |
| 25 |
empty not unset/missing. It's pointless requiring the var to be set |
| 26 |
empty in the ebuild because portage (by extension repoman) doesn't |
| 27 |
know the difference between unset and empty. |
| 28 |
|
| 29 |
Making it aware of the difference is a bit daft also imo; a package |
| 30 |
that lacks a homepage I tend to think should be flagged as a warning, |
| 31 |
regardless if the author set the HOMEPAGE var or not. |
| 32 |
~harring |
Archives