| 1 |
On Wed, 25 Nov 2009 23:59:45 +0100 |
| 2 |
Ulrich Mueller <ulm@g.o> wrote: |
| 3 |
> Real examples would be issues like bugs 83877 [1] or 263387 [2]. |
| 4 |
> Nothing that could be easily dismissed or worked around. Both issues |
| 5 |
> are fixed with Portage since a long time. |
| 6 |
|
| 7 |
Yes, those are examples of packages relying upon something that is |
| 8 |
undefined behaviour, and that behaves differently depending upon the |
| 9 |
Portage version you use. |
| 10 |
|
| 11 |
> I don't know of any example where non-preservation of nanosecond |
| 12 |
> timestamps would cause problems. |
| 13 |
|
| 14 |
Not non-preservation. Partial and inconsistent corruption. |
| 15 |
|
| 16 |
srcfile gets installed with its mtime preserved, because it is installed |
| 17 |
using rename. |
| 18 |
|
| 19 |
objfile, whose mtime should be a bit later than srctime's, gets |
| 20 |
installed with its mtime corrupted to be slightly less than it should |
| 21 |
be, and less than srctime's, because it is installed the slow way. |
| 22 |
|
| 23 |
A program checks that objfile's mtime is greater than or equal to |
| 24 |
srcfile's mtime. That check will fail, and reinstalls will randomly fix |
| 25 |
and break it depending upon which way the corruption goes. |
| 26 |
|
| 27 |
-- |
| 28 |
Ciaran McCreesh |
Archives