Bug 192714
| Summary: | [stage] mail/milter-greylist-devel - update to 4.5.11, add staging, take maintainership | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | Ports & Packages | Reporter: | Daniel Austin <freebsd-ports> | ||||||
| Component: | Individual Port(s) | Assignee: | John Marino <marino> | ||||||
| Status: | Closed FIXED | ||||||||
| Severity: | Affects Some People | CC: | marino | ||||||
| Priority: | --- | ||||||||
| Version: | Latest | ||||||||
| Hardware: | Any | ||||||||
| OS: | Any | ||||||||
| Attachments: |
|
||||||||
|
Description
Daniel Austin
2014-08-16 20:51:38 UTC
move tested staging PR to patch-ready status boom! man pages are *always* installed. Anything that has conditional installation like we see here (conditional per option MANPAGES) is flat-out wrong. actually it looks like you added that yourself. Ah, I added that to keep it in-line with the behaviour of the non-devel version of the port (mail/milter-greylist) okay, that maintainer is known for that. it's wrong. It would be helpful to submit a PR on it. secondly, RUN_DEPENDS+= BUILD_DEPENDS is also wrong. The handbook specifically says not to do that. With the options it's really confusion what the run depends is supposed to be because the options are also assigning run depends. Can you explicitly list what the run depends is supposed to be? Should I assume it the "no options set" value of build depends? tip: Never have "post-install:" target when "do-install:" target is defined. The post-install steps are then added to install target. very advanced obscure tip:
".include "${PORTSDIR}/mail/sendmail/..." is wrong
It should be:
".include "${.CURDIR}/../../mail/sendmail" or
".include "${.CURDIR}/../sendmail"
PORTSDIR is absolute path, we want relative paths.
(In reply to John Marino from comment #5) > okay, that maintainer is known for that. it's wrong. It would be helpful > to submit a PR on it. > > secondly, RUN_DEPENDS+= BUILD_DEPENDS is also wrong. The handbook > specifically says not to do that. With the options it's really confusion > what the run depends is supposed to be because the options are also > assigning run depends. > > Can you explicitly list what the run depends is supposed to be? Should I > assume it the "no options set" value of build depends? That should be a safe assumption - without options set, it only requires sendmail to be installed (base or port) for access to libmilter.so Do we really need an 800-column line in pkg-plist? 1) is the warning still needed? 2) why can't this information be in pkg-message or UPDATING? (In reply to John Marino from comment #9) > Do we really need an 800-column line in pkg-plist? > 1) is the warning still needed? > 2) why can't this information be in pkg-message or UPDATING? I can't see any reason not to put it in pkg-message. The only difference is that it's currently only shown if the old database location is detected, but i'm sure people are capable of checking for themselves! The warning can probably be removed altogether now I guess... I don't imagine many people will still have an old version installed with the old path in use. okay. Can you tell me why you are saving this port? I generally despite -devel ports. Why do we need this -AND- the original? Is the maintainer blocking the update? Is this alpha/beta or just newer? oh crap, half my changes are on the original port. another reason to hate -devel ports, they screw with tab completion. (In reply to John Marino from comment #12) > okay. > Can you tell me why you are saving this port? > I generally despite -devel ports. > > Why do we need this -AND- the original? > Is the maintainer blocking the update? > Is this alpha/beta or just newer? 4.5.x is listed as "development snapshot - not for production use" vs 4.4.x as "stable" so I guess we should keep them separate. Whether that's just the author not committing to creating a new stable release i'm not sure. There's no release of 4.5 at all? aside: INSTALL_* commands are masked with "@", that's not allowed. (In reply to John Marino from comment #15) > There's no release of 4.5 at all? nope, no 4.5.x release :-( http://hcpnet.free.fr/milter-greylist/#downloads (4.4.3 latest stable, 4.5.11 latest devel) (In reply to Daniel Austin from comment #11) > The warning can probably be removed altogether now I guess... I don't > imagine many people will still have an old version installed with the old > path in use. Is 4.4 "the old version" or does this pertain to a much older release? (In reply to Daniel Austin from comment #8) > (In reply to John Marino from comment #5) > > Can you explicitly list what the run depends is supposed to be? Should I > > assume it the "no options set" value of build depends? > > That should be a safe assumption - without options set, it only requires > sendmail to be installed (base or port) for access to libmilter.so Ironically, BUILD_DEPENDS is empty in the base case. I ran remove the line completely. (In reply to John Marino from comment #18) > (In reply to Daniel Austin from comment #11) > > The warning can probably be removed altogether now I guess... I don't > > imagine many people will still have an old version installed with the old > > path in use. > > Is 4.4 "the old version" or does this pertain to a much older release? Release date for 4.4.3 is 7th March 2013. 4.5.1 build date is 20th May 2013. I'm assuming the author doesn't want to commit to pushing 4.5 to stable. 11 builds over just under a year within 4.5.x branch. But, i'm also weary that milter-greylist is quite a popular installation with FreeBSD users. I'm just trying to decide if that warning is still necessary. I can't figure that out from what you've told me. However, I think "pkg-message" is fine instead of trying to detect it. okay, it busted on config. I'm going to upload the current diff and the partial build log, maybe you can spot the issue. Created attachment 145911 [details] reworked version breaks here: checking if compiler accepts -Wall... yes checking if compiler accepts -Werror... no checking if ld accepts -R... yes checking if ld accepts --rpath... yes checking for ldap_search in -lc -lldap_r-2.4 -llber-2.4... no checking for ldap_search in -lc -lldap_r -llber... no Error: no OpenLDAP library names located! ===> Script "configure" failed unexpectedly. Please report the problem to freebsd-ports@dan.me.uk [maintainer] and attach the "/wrkdirs/usr/ports/mail/milter-greylist-devel/work/milter-greylist-4.5.11/conf including the output of the failure of your make command. Also, it might be a good idea to provide an overview of all packages installed on your system (e.g. a /usr/local/sbin/pkg-static info -g -Ea). *** Error code 1 it's explicitly configured without openldap support, so I don't know why it's checking for it. I'll look at config.log now. (In reply to John Marino from comment #24) > it's explicitly configured without openldap support, so I don't know why > it's checking for it. I'll look at config.log now. interesting... if you don't put "--without-openldap" to configure, it configures fine (and without openldap) i was thinking it was a bug in configure, that confirms it. I'll put it back to how it was with a comment. okay, it builds and passes poudriere like that. We have one final thing to decide: Should we maintain the warning? If you say yes, I'll add it to pkg-message.in. If you say no, I'll commit it as it is. (In reply to John Marino from comment #27) > okay, it builds and passes poudriere like that. > > We have one final thing to decide: > > Should we maintain the warning? > If you say yes, I'll add it to pkg-message.in. > If you say no, I'll commit it as it is. Scrolling through commits for Makefile, that warning has been there over 5 years. Based on that, i'm very happy for it to be removed completely. It shouldn't be affecting anyone now. (In reply to Daniel Austin from comment #28) > Scrolling through commits for Makefile, that warning has been there over 5 > years. > Based on that, i'm very happy for it to be removed completely. > It shouldn't be affecting anyone now. Hypothetically, if somebody stayed on milter-greylist (not devel) and moves to devel today, would they need to heed the warning? (In reply to John Marino from comment #29) > (In reply to Daniel Austin from comment #28) > > Scrolling through commits for Makefile, that warning has been there over 5 > > years. > > Based on that, i'm very happy for it to be removed completely. > > It shouldn't be affecting anyone now. > > Hypothetically, if somebody stayed on milter-greylist (not devel) and moves > to devel today, would they need to heed the warning? No, both use the same path for the file. And the warning on non-devel port has been there >5 years also. A commit references this bug: Author: marino Date: Sun Aug 17 11:08:06 UTC 2014 New revision: 365171 URL: http://svnweb.freebsd.org/changeset/ports/365171 Log: Stage mail/milter-greylist-devel and upgrade version 4.5.7 => 4.5.11 Also assign maintainership to submitter PR: 192714 Submitted by: Daniel Austin Add'l work by: marino Changes: head/mail/milter-greylist-devel/Makefile head/mail/milter-greylist-devel/distinfo head/mail/milter-greylist-devel/files/patch-configure head/mail/milter-greylist-devel/pkg-plist it would be really great if you could make an new PR for milter-greylist that makes all the same changes except don't touch the warnings. As long as it behaves exactly the same (except for manpages) then I can commit it without getting approval first. (In reply to John Marino from comment #32) > it would be really great if you could make an new PR for milter-greylist > that makes all the same changes except don't touch the warnings. > > As long as it behaves exactly the same (except for manpages) then I can > commit it without getting approval first. Done - #192740 |
