| 1 |
On 06/26/2011 07:55 PM, Matt Turner wrote: |
| 2 |
> I claim that |
| 3 |
> (1) app-text/asciidoc and its dependencies are not unreasonable for |
| 4 |
> catalyst-9999 |
| 5 |
> (2) app-text/asciidoc and its dependencies should be avoided for |
| 6 |
> catalyst snapshots/releases |
| 7 |
> (3) we can simply modify the timestamp of the generated files to be |
| 8 |
> that of the most recent commit that `git archive` has access to, |
| 9 |
> thereby allowing us to reproduce identical tarballs |
| 10 |
|
| 11 |
I played with this in the Makefile now. The only easy solution that |
| 12 |
really gives stable checksums is setting time including modification |
| 13 |
time to the time of the related commit. So that means loss of real |
| 14 |
modification times. If you want to go that route the command |
| 15 |
|
| 16 |
touch --date=$(git log --pretty=%ci -1) foo |
| 17 |
|
| 18 |
could be handy. I guess git archive looks at timestamps from the git |
| 19 |
store only to avoid that. Using git archive and pulling generated files |
| 20 |
in after may work, but it#s more research than I would like to do myself. |
| 21 |
|
| 22 |
|
| 23 |
> (4) checking in generated content into git is dirty. We should not do this |
| 24 |
> |
| 25 |
> Do you disagree with any of these points, and if so, which? |
| 26 |
> |
| 27 |
> Matt |
| 28 |
> |
Archives