1 |
On Sun, Nov 16, 2008 at 10:05 AM, Zac Medico <zmedico@g.o> wrote: |
2 |
> -----BEGIN PGP SIGNED MESSAGE----- |
3 |
> Hash: SHA1 |
4 |
> |
5 |
> Alec Warner wrote: |
6 |
>> [1] I fully support server-side tests but I have low expectations for |
7 |
>> their implementation or deployment. |
8 |
> |
9 |
> Why so low? I'm pretty optimistic. ;) |
10 |
|
11 |
Because the only person willing to work on them is you, and you are a |
12 |
single point of failure; + busy. |
13 |
|
14 |
> |
15 |
>> Also it stands to reason that |
16 |
>> distributed tests scale better in the long run and it is likely that |
17 |
>> we could make our local tests run much faster if we dedicated some |
18 |
>> developer time to it. I am sensing a possible SOC project here? |
19 |
> |
20 |
> The problem with running the tests locally is that our dataset is |
21 |
> pretty large and the types of QA tests that it needs are pretty |
22 |
> expensive. Given these constraints, it makes more sense to do the |
23 |
> tests on a dedicated server than to do it on a client which may have |
24 |
> limited resources. |
25 |
|
26 |
Well if you run the expensive tests on our large datasets every time |
27 |
someone commits you may soon destroy the scm server too without some |
28 |
modifications, either to the tests (a daemon that caches tree state?) |
29 |
or to our commit processes (pushing less often means tests run less |
30 |
often means server doesn't die) or some combination of these things. |
31 |
|
32 |
If the entire git tree is in a memory-backed FS; I'm not sure how fast |
33 |
the checks are (if we eschew the memory-based daemon thing). |
34 |
|
35 |
> - -- |
36 |
> Thanks, |
37 |
> Zac |
38 |
> -----BEGIN PGP SIGNATURE----- |
39 |
> Version: GnuPG v2.0.9 (GNU/Linux) |
40 |
> |
41 |
> iEYEARECAAYFAkkgYPMACgkQ/ejvha5XGaONpQCeJRZY1JHxreTbvQpi8R9PIlDh |
42 |
> 6EYAoKEbAfmp5r5XplG1JhU+zMMK+jyp |
43 |
> =R44q |
44 |
> -----END PGP SIGNATURE----- |
45 |
> |