Suggestions several

[walterar]

The file name of the latest nightly builds binary is somewhat confusing.

I suggest using x for the year, for the month xx, xx for the day, xx for the daily version

Ex 31010.01 for the first compilation of today, etc.

Or any other manner that is consistent and intuitive.

[bszlrd]

Isn't the current one yyyymm.counter? I personally don't find it confusing. I usually think omitting the first two digits of the year could be more ambiguous. Maybe yyyy.mmdd.counter?

[walterar]

nozmajner, look at the last night and you will realize what I mean.

http://sourceforge.net/projects/pioneerspacesim/files/

[walterar]

pioneer-201310.107-win32.7z
pioneer-201310.117-win32.7z

If you pick the latest version, which you choose you?

[bszlrd]

Oh, I see your point now. I never noticed that.

[walterar]

Regarding: 2013 to 3

With 3 serve until 2019

But it is not relevant, you can add the century if you want. What happens is that I come from a time where I had to economize bytes :)

This is not a serious problem, it is actually very simple to fix in the script, and robn surely know how to do, if find five minutes of time.

[jpab]

It looks like the script that picks the build version number is broken. The version is supposed to be YYYYMM.BB where YYYY is the year, MM is the month, and BB is some kind of count of commits so far that month. I don't know how that commit count is being computed, but it's clearly incorrect because it should only ever increase within one month. This month we've had (most recent to least recent): 107, 117, 72, 86, 77, 55, 10, 47, which doesn't make any sense at all.

Anyway, I think we should keep the full four digit year because it's easier then to see that the version number is based on the date, but I don't have any objection to including the day too. Either way, the final counter value needs to be fixed somehow.

[robn]

The count has always broken, and I still don't really understand why. Its really simple, one command to count all the commits since the start of the month:

Code: Select all

git rev-list --since=2013-10-11 --count HEAD

But it seems to be quite inconsistent. I've seen it produce different results on different versions of git and on different platforms. I suspect it has problems with merge commits, and with commits around the end/start of the month, perhaps something to do with timezones. But I mostly didn't care because I figured it could never go backwards, right? And yet it does.

So screw it. I'm just going to switch to YYYYMMDD, and there won't be a count. The downside is that I can't do more than one build in a day, but that's pretty rare anyway.