svn commit: r416256 - head/emulators/i386-wine-devel
Alexey Dokuchaev
danfe at FreeBSD.org
Thu Jun 2 09:45:05 UTC 2016
On Thu, Jun 02, 2016 at 01:29:12AM +0200, Mathieu Arnold wrote:
> This is completely bogus, you can't have more than one TIMESTAMP line, how
> did you manage to generate that file ?
I've tried to warn you that these things will happen, but you won't listen.
OK, let's try again.
- Unattached TIMESTAMP is bogus as it is not clear what distfile it is
related to. (If it does not relate to any of them, it should not be part
of distinfo.) No, you cannot expect that your code is flawless or that
committers won't have their own procedures to generate/maintain distinfo;
- It should not be set by committer: this is error prone and it is possible
to set it deliberately to some wrong value (e.g. 0) which might break
things in a weird way. Even when there's no ill will involved, it is not
wise to trust local clock as that might not be accurate: two commits in a
short time window might result in TIMESTAMP value go *backwards* because
of untuned clock, DST shift not applied correctly, etc.
Bottom line: it should be set by the SVN, server-side, and not mix this with
other data. Baptiste said keywords is not an option because we officially
support broken VCS that does not support them (Git), but in this case IMO it
could be emulated (and thus stored in expanded form) via commit hook (simply
replace first line with a new one; line should be commented out and maybe
marked as ``# $Timestamp: $'' to look like real keyword and help to avoid
messing with real data line SIZE() and SHA256() values).
./danfe
More information about the svn-ports-head
mailing list