| 1 |
Mikael Hallendal wrote: [Mon Jan 14 2002, 5:16:02AM EST] |
| 2 |
> mån 2001-11-12 klockan 03.52 skrev Aron Griffis: |
| 3 |
> > jano wrote: [Sun Nov 11 2001, 09:49:49PM EST] |
| 4 |
> > > I don't understand why this works for some and not for others; is it an |
| 5 |
> > > issue of system performance? |
| 6 |
> > |
| 7 |
> > It's a race condition, and it should never be done. Instead, ebuild |
| 8 |
> > authors should use dosed, which is exactly for this purpose. |
| 9 |
> > |
| 10 |
> > dosed 's/foo/bar/' Makefile |
| 11 |
> |
| 12 |
> Is this supposed to work? |
| 13 |
> |
| 14 |
> It doesn't do anything for me. |
| 15 |
|
| 16 |
dosed is currently written to only work on files in ${D}, and the sed |
| 17 |
script defaults to deleting all occurances of ${D} in the named files. |
| 18 |
This makes it ideal for "fixing" paths in text files during the ebuild |
| 19 |
'install' phase, but useless for patching sources during the 'unpack' |
| 20 |
phase. |
| 21 |
|
| 22 |
Not that you would know this from reading the dosed internal |
| 23 |
documentation, since there isn't any. :-/ |
| 24 |
|
| 25 |
Anyway, perhaps someone should update dosed or writing something similar |
| 26 |
for use during 'unpack'. Until then, what I generally do is: |
| 27 |
|
| 28 |
t=Makefile |
| 29 |
cp $t $t.orig |
| 30 |
sed 's/foo/bar/' $t.orig > $t |
| 31 |
|
| 32 |
--Chouser |
Archives