Systems manager and facilitator for Research Computing at the University of Wisconsin -- Milwaukee
- User Since
- Nov 27 2017, 4:37 PM (7 w, 5 d)
Wed, Jan 17
Apparently this *is* by design:
I think you've discovered a BSD make bug. Using -j should not change the results of the build no matter how the makefile is coded.
Thu, Dec 28
Yeah, I realize I forgot (mentor) right after I hit enter..
Fri, Dec 22
Good catch on the +=, that was an oversight.
Dec 22 2017
Added that to my checklist
Dec 21 2017
Dec 18 2017
So I got an email at 8:46pm stating that the revision was accepted, but the commit message stated that it was in a needs review state. Guessing that's because I restored the svn tag line to match the previous commit and forgot to do an arc update after. My working copy contained $FreeBSD$ and I was afraid the commit might have problems if this line differed from the previous commit. Based on what I've seen tonight, I'm thinking the old content is ignored and the new tag is inserted in its place.
Got it, thanks.
Oh, that line... I thought you were referring to $FreeBSD$ not showing the svn tag. I apparently misinterpreted our earlier discussion as a call to remove the "Created by" comments. I'll leave it in for this commit, pending clarification.
I'm not clear on what you mean, but apparently "svn diff" does not include the tag in its output if that's what you're getting at. I never noticed this before, though. I just home in on the changes.
Dec 17 2017
Dec 15 2017
Add PORTREVISION, silence RM
I'd lean toward echoing STRIP_CMD, so I'd amend that to
Perhaps there could be agreement on a general principal regarding what should be echoed?
No hurry on my end, especially since we have a working port already committed. Let's use this as an opportunity to learn and maybe improve the docs before it's committed.
Yeah, feels like the same revision since I reopened the differential. Too much to keep track of for my worn-out brain...
Remove redundant conf file from EXAMPLES_DIR
My instinct said to stick with the canonical method (STRIP_CMD), now I know why. Thanks...
Dec 14 2017
It *is* staged, so the build system need not worry about it, but the ports install target works from the plist and ignores whatever is commented out.
Yes, they are the same.
Add DOCS and EXAMPLES to OPTIONS
One more patch needed
What's the traditional approach when arc diff --update produces this?
The first time you mentioned it, I suggested removing PORTDOCS and PORTEXAMPLES from options because they only install 4 very small files. I did not see a response to that suggestion.
I removed them from the Makefile, which I thought was the issue. Are you referring to the %%PORTDOCS%% and %%PORTEXAMPLES%% in pkg-plist? If so, I can remove them as well. They were generated by makeplist.
Dec 13 2017
Thanks for all the feedback, guys. This one is out of my usual realm as well. Most of my ports are scientific apps running on headless servers...
Fix config file issue. Awaiting feedback on other items...
- Porter's Handbook section 184.108.40.206. Syntax has this:
Think I've got it now...
Dec 12 2017
I think 4 months without a response to Tobias is enough. Just wanted to know if there are any strict rules about this situation. I'm going to let the last of 4 poudriere builds finish and then I'll do the commit assuming it succeeds.
Haven't heard from the maintainer yet. He submitted part of the patch on Aprll 6th and that's the last anyone heard from him. Should I try to reach out to him in case he didn't see the bugtracker updates?
Add python flavor to RUN_DEPENDS
I'm getting swamped again this morning, so I'll get back to these later when I can focus again...
NLS is on by default, so I'm thinking we need to explicitly disable it for the option to have any effect:
Add Python flavor and silence REINPLACE_CMD.
Dec 9 2017
Rename with caps to match PORTNAME.
Dec 8 2017
Alrighty, so I have a powerful ally...
Dec 7 2017
I think mat's comment about python was confirming that USES=python:env is correct, but python:run wouldn't do any harm in this case. Since this port doesn't use python, but needs PYTHON_PKGNAMEPREFIX and PY_FLAVOR, I think python:env is the best pick.
Good stuff, thanks! I reran it through poudriere just to be safe.
Change PORTNAME to match upstream dist
Thanks for clarifying the python dependency issue. I wasn't sure about that one... I updated my checklist accordingly.
Actually, parens would have no effect in this case. The cd only affects other commands on the same line anyway, since each line is executed under a new shell process that's a direct child of the make process. The Makefile below will show the different shell processes for each line and different output from pwd:
Fix issues noted by mentors
Thanks for the tips. Can't believe I missed DISTVERSION... ;-)
Dec 6 2017
Add missing USES=python arg
Dec 5 2017
Thanks, I hadn't given it a full review since commandeering...
Thanks for the tips. I hadn't thought to use bulk -t instead of testport. My plan was to manually run testport again with @py36 when needed. I see that numpy gets the VIP treatment in python.mk. I'll keep an eye on that section for additions in the future...
Dec 4 2017
Add Makefile. Will send upstream along with malloc patch.
Fix core dump issue. Code was assuming malloc() aligns to 16 byte boundaries. May be a Linuxism.
Remove wip category
Update to new release 2.1.10.
Dec 3 2017
I don't think we have anyone using fasttree presently, but if you send me a test case that produces the core dump, I'll try and reproduce it.
Dec 2 2017
Settled on a usable PORTVERSION
I agree with PORTVERSION=2.1.0.p20170505. That would have been the next thing I came up with myself, all seemingly reasonable alternatives eliminated.