| 1 |
-----BEGIN PGP SIGNED MESSAGE----- |
| 2 |
Hash: SHA1 |
| 3 |
|
| 4 |
Ciaran McCreesh wrote: |
| 5 |
> On Tue, 05 Aug 2008 22:55:06 -0700 |
| 6 |
> Zac Medico <zmedico@g.o> wrote: |
| 7 |
>> Would having the ebuild perform locking be unportable or introduce |
| 8 |
>> any undesirable complexity? Does it really need to be so |
| 9 |
>> fine-grained? |
| 10 |
> |
| 11 |
> It's not particularly tricky, and the lock code can be moved into an |
| 12 |
> eclass if necessary. And the gain is worth it -- scm fetches can be |
| 13 |
> slow, so parallelising fetches from different hosts will speed things |
| 14 |
> up quite a bit. |
| 15 |
> |
| 16 |
>> I don't see flock listed in the spec here [1]. |
| 17 |
> |
| 18 |
> If any platform doesn't have flock, there're various other ways of |
| 19 |
> doing locking using mkdir and either a fifo or just plain old sleep. |
| 20 |
> |
| 21 |
|
| 22 |
To simplify things, how about if we just do a |
| 23 |
PROPERTIES=live-src-unpack for now, to indicate exclusive access to |
| 24 |
$DISTDIR during src_unpack? Thats a simple and portable baseline |
| 25 |
that will be quite useful even without anything finer grained. |
| 26 |
|
| 27 |
Zac |
| 28 |
-----BEGIN PGP SIGNATURE----- |
| 29 |
Version: GnuPG v2.0.9 (GNU/Linux) |
| 30 |
|
| 31 |
iEYEARECAAYFAkiZRA0ACgkQ/ejvha5XGaNZsACfVMIzQgQufKgbMrM6TyjKowN4 |
| 32 |
PQIAn1jTnV7omQaoezbU3cR4eS/hZrAU |
| 33 |
=I00K |
| 34 |
-----END PGP SIGNATURE----- |
Archives