[Development] Reproducible Qt ; the purpose of QLibraryInfo::buildDate

[Alejandro Exojo]

El Sunday 15 February 2015, Sune Vuorela escribió:
> On 2015-02-15, Alejandro Exojo <suy at badopi.org> wrote:
> > People might rely on the function in proprietary applications in ways
> > that are impossible to predict.
> 
> Every time we fix a bug, we might introduce regressions for people
> relying on the bugs. That's not a reason for not fixing a bug.

I fail to see how changing QLibraryInfo::buildDate to return a hardcoded 
string instead of the intended build date is fixing a bug. Is obvious that you 
are breaking a feature in order to gain a different one that you think is more 
valuable.

I could agree with you on this, but not at the cost of breaking a previous 
feature in a minor release of Qt. It worries me a lot to see this kind of 
changes in minor releases. It sets a precedent.

Not only that. I still think that you could have the reproducible builds 
feature without introducing breakage: just fake the date. Hardcode the date 
where you care about the issue, and not for everybody else. I see there is 
something already available on Debian:

https://packages.debian.org/sid/sdate

-- 
Alex (a.k.a. suy) | GPG ID 0x0B8B0BC2
http://barnacity.net/ | http://disperso.net