Created attachment 161867 [details] patch The attached patch fixes the build problem (failure message suggests to add the shebang location).
Created attachment 161869 [details] poudriere log
You haven't explained the problem. (you mention "the build problem" but haven't established what the problem is). Secondly, you're referencing bash but I don't see any build depends or run depends on bash. I'm afraid the introduction to this PR is lacking. e.g. 1. what's the problem 2. suggestioned solution (the patch) 3. explain why the patch solves the problem 4. impact of not applying the patch (i.e. rejecting the solution) The portmon is all green lights so I'm not sure what this problem is.
ah, i see the build depends on bash now. I missed that at first. still, what build error are we trying to solve and what happens as a result of this error?
Sorry I didn't properly explain the problem in the OP. Here is the build error log I am getting: > ====> Compressing man pages (compress-man) > ====> Running Q/A tests (stage-qa) > Error: '/bin/bash' is an invalid shebang you need USES=shebangfix for 'bin/autoopts-config' > *** Error code 1 The port already has bash as a runtime dependency. I am not sure why you aren't seeing the same error, because work/autogen-5.18.4/autoopts/autoopts-config does have the wrong shebang in it. 10.2-STABLE
so why are you expecting anyone that peruses this PR to be aware of a stage QA error? How would we know this? Other than building the port in poudriere with full testing, check the log, and then look back at the PR. I would have pasted that excerpt of the log into the PR. These stage-QA errors don't break standard builds because they are extra testing that isn't done on the cluster.
btw, bash is NOT a run depends. Should it be? Is this autoopts-config script a core executable? if so, bash should be a run depends.
I got the error through the simple 'make' command. Not sure why stage-qa runs through the make command for me. But it isn't an entirely bad idea to build in poudriere with stage QA. > I would have pasted that excerpt of the log into the PR. I just did, didn't I?
more explicit phrasing: "I would have pasted an excerpt of the log in the description of the PR to establish the problem". It's not happening to everyone because stage-qa is not run by default. you must have developer mode or something that is triggering it. the missing run-depends (if it's missing) will require a new patch.
> btw, bash is NOT a run depends. > Should it be? Is this autoopts-config script a core executable? if so, bash should be a run depends. I am not particularly familiar with autogen, but I don't see any installed by it exacutables using bash. So why would it need to be the runtime dependency? Ah yes, I had developer mode "on", that's why I even got the error. So no it is all clear, right?
Ah, okay, it does need runtime dep. I will re-submit the patch
Created attachment 162130 [details] Updated patch Updated patch addresses these issues: 1. It fixes the above-mentioned shebang problem (message during the stage QA) 2. Fixed the port comment (doesn't begin with "The" now) 3. Added the license clause
A commit references this bug: Author: marino Date: Fri Oct 16 21:25:36 UTC 2015 New revision: 399523 URL: https://svnweb.freebsd.org/changeset/ports/399523 Log: devel/autogen: Fix stage-qa failure The autoopts-config script needs bash, so it's been added as a run dependences and the shebang wsa fixed. While here, add the GPLv2 LICENSE and improve the COMMENT. PR: 203672 Submitted by: Yuri (rawbw.com) Changes: head/devel/autogen/Makefile
thanks