Re: HEADS UP: Merge of binutils 2.17

On Jan 07, 2011, at 17:37 , Doug Barton wrote:
> On 01/07/2011 13:54, Ade Lovett wrote:
>> 
>> Most likely it's low priority given all the other exp-runs that
>> affect 7.x/8.x, tweaking things for an 6.x-EOL-tagged tree, and a
>> bunch of other infrastructure stuff.  Not to mention the impending 7-
>> and 8- RELEASEs.

Before I start on this, I would like a few things noted for the record:

1.  I have set Reply-To to developers_at_ (this should be a major hint)
2.  I am not a current member of portmgr_at_
3.  I requested, and served, for a very short time, on the first portmgr
  

> That may very well be the case, but if so then it's incumbent on portmgr to communicate that. If you check the audit trail you will find that they did not.

Horsecrap.  You are taking an individual PR history without reference to the whole host of things that were also going on at the same time.  Like it or not, when it comes to ports, -STABLE wins over -CURRENT every single time.

> IMO this is a total red herring, and has been for several years now. I run -current every day on my real-work system, and barring the occasional hiccup it's been buildable nearly every time I've tried.

Apologies for not being able to drive my email client appropriately.  The issue at hand is one of running -CURRENT.

There is a distinct, and fundamental difference between running -CURRENT on a single system, as opposed to a cluster of systems that are tightly interlinked.   I do not doubt that -CURRENT works for you on your individual machines.  If you would like a taste of how heavily package build clusters stress out whatever host system they are running on, then I urge you to install one of the two tinderbox ports under ports-mgmt, proceed to add, let's say, x11/gnome2 or x11/kde4, and run the build.

make buildworld/buildkernel/installworld/installkernel plus associated steps is in fact an exceptionally tiny subset of what FreeBSD actually does on a daily basis.  Even more so when it comes to the bulk building of packages that apparently a lot of folks rely on.

> The way I would approach the problem of building packages for -current is to pick a day to update the src tree, then do the following:

Sadly, the only thing I can say to your 4-step procedure, and with utmost politeness, is that your src-centric views are completely missing the point.  "4. start building ports" is in fact a 20- or 30-step process to ensure no cross-contamination.  Even a cursory glance at /usr/ports/Tools/portbuild would verify this.  No-one really likes having massive clusters, requiring continual attention (hardware failures and so on).  Really.

> But the current system of "don't do anything" just isn't cutting it.

I look forward to your input and total solutions on how to make this better.  I do.  This may sound sarcastic, but I am absolutely, positively, 100-percent looking for better solutions, particularly in situations where, to take a random example, the entire existing compiler base is removed and replaced with something better.

Doug, you have comprehensively shown that in its current (sic) instantiation, the package building cluster is completely, utterly, and totally incapable of keeping up with the sandbox that is -CURRENT.

I for one look forward to your proposed solutions to this righteous problem.

Regards,
-aDe