1 |
>>>>> On Mon, 27 Feb 2017, Mike Gilbert wrote: |
2 |
|
3 |
>> d) Should git expansion of $Id$ be enabled (i.e., ident in git |
4 |
>> attributes)? |
5 |
|
6 |
> If $Id$ is to be kept, I think point "d" needs further clarification: |
7 |
> in what contexts should ident expansion be enabled? |
8 |
|
9 |
> At rsync generation time? |
10 |
|
11 |
> In the development repo? (via .gitattributes). |
12 |
|
13 |
> Enabling expansion in some places but not others may cause some issues |
14 |
> that would necessitate further work to prevent "$Id: xxxxxxx $" |
15 |
> strings from being committed in the development repo accidentally. |
16 |
> This might mean a repoman check and/or a git hook. |
17 |
|
18 |
Right, there are many paths to insanity there. That's why we should get |
19 |
rid of the whole thing altogether. |
20 |
|
21 |
A content tracker should faithfully keep track of the files committed, |
22 |
but not modify their contents. Even CVS didn't reduce $Id: something$ |
23 |
to $Id$, but kept the original version in the repository [1]. |
24 |
|
25 |
Ulrich |
26 |
|
27 |
[1] http://cvsman.com/cvs-1.12.12/cvs_99.php#SEC99 |