Date: Thu, 19 Mar 1998 01:23:37 +0100 From: Eivind Eklund <eivind@yes.no> To: "Steven G. Kargl" <kargl@troutmask.apl.washington.edu>, freebsd-ports@FreeBSD.ORG Subject: Re: bsd.port.mk bug??? Message-ID: <19980319012337.39204@follo.net> In-Reply-To: <199803190011.QAA15916@troutmask.apl.washington.edu>; from Steven G. Kargl on Wed, Mar 18, 1998 at 04:11:36PM -0800 References: <19980319000854.30236@follo.net> <199803190011.QAA15916@troutmask.apl.washington.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Mar 18, 1998 at 04:11:36PM -0800, Steven G. Kargl wrote: > According to Eivind Eklund: > > On Wed, Mar 18, 1998 at 11:48:58AM -0800, Steven G. Kargl wrote: > > > > Let's say checking a port for problems with obsolete TK versions take 3 > > minutes on average (and some of the cases are pretty complex, so I don't > > think that's runreasonable). 1400*3 = 7000 minutes, approximately 3 work > > weeks. > > > > It's Too Expensive To Check. > > > > Was that simple enough words for you? > > > > BTW: Looking at dependencies is not good enough, just to counter that > > argument before it shows up. > > > > I don't see any smiley faces, so I'll assume you are serious. > > There are 1078 Makefiles in my /usr/ports. I've conceded that I > don't have the entire ports trees because japanese, korean, etc > are irrelevant to me. Of those 1078 Makefiles, there are 59 > references to the string "tk" in the CATEGORIES variable. There can be > only 59 ports that have a problem with an obsolete tk.h. Damn, the dependencies/categories-argument. The one I told you didn't hold. OK, I'll go through the details: This isn't enough - there are ports that build conditionally on whether things are present, even though they don't bring them in if they aren't present. We don't know if any does that for TCL or TK, but it is likely enough to require that every port be checked. And this isn't actually a cost which any of us are at all interested in bearing. Thus, "the quick way out" - force people with bugged systems to remove those header files. Yes, it is correct that this wouldn't be the ultimate solution if we had infinite manpower. However, we don't, so that's the solution we use. The users get a more consistent system as an added benefit. If you are prepared to go build all the ports with and without all the three different TK-cases so you can add a 'builds OK with buggy TK' variable to each of them - feel free to. Here are a few issues you'll have to contend with: You'll have to check that no extra files are installed when the build is done with the wrong stuff installed. You'll have to check that each port package equally (to the byte) with and without. Some ports are interactive; you'll have to do them by hand. Some ports are presently broken - you have to check them when they get fixed again. Some ports are not yet added - you'll have to check them as they are added. Some ports are incompatible - they can't be in the same system at the same time. You have to solve this (by being careful what you build when). You can probably automate much of this, but the parts you can't automate is going to be damn expensive. I'd suggest it would probably be more productive to use that time for something else. An improvement to the text in bsd.port.mk that tripped you would be a good start ;-) Eivind. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-ports" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19980319012337.39204>