1 |
On Sun, Nov 16, 2008 at 9:20 AM, Donnie Berkholz <dberkholz@g.o> wrote: |
2 |
> On 01:34 Sun 16 Nov , Alec Warner wrote: |
3 |
>> Change the workflow[1]. |
4 |
>> |
5 |
>> Right now you pay the cost of running tree-wide tests every time you |
6 |
>> commit. This incentivizes developers to commit less often (to avoid |
7 |
>> paying the tax of tree-wide tests). |
8 |
>> |
9 |
>> In CVS commiting less is a problem: |
10 |
>> |
11 |
>> 1. Make Changes to a number of files. |
12 |
>> 2. Script your commits. |
13 |
>> 3. Run Script |
14 |
>> 4....N: Script commits one file at a time. |
15 |
>> 4.5: A race condition between changes you have commited to CVS versus |
16 |
>> uncommited changes occurs when CVS is synced to Osprey. This race |
17 |
>> condition can often cause tree oddness. |
18 |
>> 5. All Changes commited. |
19 |
>> 5.5: All changes synced to Osprey. |
20 |
>> |
21 |
>> I am unsure if repoman category and repoman tree-wide commits avoid |
22 |
>> this race condition. |
23 |
>> |
24 |
>> A new scheme would be: |
25 |
>> |
26 |
>> 1. Make Changes to a number of files. |
27 |
>> 2. Category or Treewide Repoman commit. |
28 |
>> 3. Run taxing tree-wide tests. |
29 |
>> 4. git commit -a (?) |
30 |
>> 5. Done! |
31 |
>> |
32 |
>> Tell me if I'm horribly wrong or missing something, it is late here. |
33 |
> |
34 |
> Are you saying we could use the index here? Another option (that you may |
35 |
> or may not be suggesting) could be to make a new 'repoman push' |
36 |
> operation that actually does the hard checks, and make 'repoman commit' |
37 |
> just do a local commit and only very quick checks. |
38 |
|
39 |
That was what I was thinking. |
40 |
|
41 |
> |
42 |
> -- |
43 |
> Thanks, |
44 |
> Donnie |
45 |
> |
46 |
> Donnie Berkholz |
47 |
> Developer, Gentoo Linux |
48 |
> Blog: http://dberkholz.wordpress.com |
49 |
> |