Re: [Libusb-devel] libusb/io.c references hash in stuge?

[Pete B. <pe...@ak...>]

On 2011.09.14 13:42, Peter Stuge wrote:
> The couple of commits that went into
> libusb.git yesterday were to test that the git->Trac integration is
> still working correctly, so I pushed carefully.

If that's just a limited test, then that's fine. The problem I have is 
that I've seen a few similar pushes happen in -stuge, where the 
timestamp (from gitweb) bares very little relation with the actual time 
the commit was made public. Personally, I find a discrepancy of more 
than 48 hours between a commit and its public (gitweb) apparition into a 
repo very annoying, as, unless you pull on a very regular basis or have 
a very good memory, you can not rely on gitweb to tell you that some 
change has occurred recently.

>> in which case we have another problem).
>
> That problem (as has been said over and over) seems to be that I have
> been doing work for the last year exclusively in libusb-stuge.git
> while everyone would really prefer if this happened in libusb.git for
> more visibility. I think the testing branch is a good idea and I'll
> start using that.

That would be good. If one repo is meant to be used for testing, we want 
to avertise it as explicitly as we can. But I hope you will try to 
ensure that even if a commit has been in testing for more than a few 
days, and even if it is brought into mainline as-is, the committer 
timestamp gets updated. Again, the idea is that somebody monitoring 
gitweb to find the latest changes before issuing a pull can find out 
more easily about recent stuff.

> The Trac timeline http://libusb.org/timeline is maybe easier to
> follow, but please don't everyone click that link at once now. ;)

Nice! Don't know if you mentioned that one before but that's actually a 
great tool, and either I had forgotten about it or missed it first time 
around. Thanks for the link.


By the way, if you're picking up on the 1.0.9 release effort, I've 
recently applied a few patches to my branch that you may want to have in 
mainline: Some beautification of the MinGW/cygwin build process [1], 
removal of warnings when logging is disabled [2] and removal of a 
longstanding MinGW warning [3] (which I doubt you're gonna like as that 
one redefines timercmp in core and you will of course prefer MinGW to 
fix it on their end).

I'm assuming that I still have plenty of time before 1.0.9 to retrofit 
those against stuge/testing, so it'll probably be one week or two before 
I send them your way. But if you think 1.0.9 will happen sooner than 
that, let me know.

Regards,

/Pete

[1] 
http://git.libusb.org/?p=libusb-pbatard.git;a=commitdiff;h=3209815810a7e1deb6cb961a4dac01b9cfb64d7d;js=1
[2] 
http://git.libusb.org/?p=libusb-pbatard.git;a=commitdiff;h=52925893448df74e895d756c3280af15494d4166;js=1
[3] 
http://git.libusb.org/?p=libusb-pbatard.git;a=commitdiff;h=5f950c84feff96c942ec63bd3b0041f944e8fefc;js=1