Build Reproducibility Workshop Report

[Ryan Schmidt]

> On Dec 10, 2015, at 17:53, Clemens Lang <cal at macports.org> wrote:
> 
>> On Thu, Dec 10, 2015 at 05:39:47PM -0600, Ryan Schmidt wrote:
>> But it thought we were saying the portindex wouldn't be generated
>> correctly if use-commit-times=yes. I certainly remember problems along
>> those lines years ago...
> 
> The portindex generation re-indexes all Portfiles whose mtime is newer
> than that of the old portindex. After generation, it sets the mtime of
> the new portindex to largest mtime of any of the indexed Portfiles.
> 
> Since the newest date a portindex can have with this algorithm is the
> date of the 'svn up' and all committed changes up to this time have been
> included in the portindex, this is safe.
> 
> It is possible that the portindex generation did not always have this
> method to determine the mtime of the portindex, which would explain the
> problems you recall.

I may have been remembering

https://trac.macports.org/ticket/29223#comment:5

which was not a portindex problem and which was fixed.