mail/opensmtpd-extras-* : unconditionally define PORTREVISION to skip unneccessary bump slaves by master port touching
Details
Diff Detail
- Repository
- rP FreeBSD ports repository
- Lint
Lint Not Applicable - Unit
Tests Not Applicable
Event Timeline
opensmtpd-extras-filter-dnsbl/Makefile | ||
---|---|---|
9 ↗ | (On Diff #16994) | Your other patch removes the mail/opensmtpd dependency, and this patch removes the dns/libasr dependency. This filter requires libasr so you'll have to leave one or the other in. |
28.05.16 1:30, adamw (Adam Weinberger) пишет:
LIB_DEPENDS= libasr.so:dns/libasr
Your other patch removes the mail/opensmtpd dependency, and this patch
removes the dns/libasr dependency.
Please take a second look at patch for masterfile.
Instead of hardcoded mail/opensmtpd and dns/libasr it now configurable:
master port now have option to select between stable/devel opensmtpd,
and both opensmtpd ports have dependency via options between two
versions of libasr
This filter requires libasr so you'll have to leave one or the other in.
BTW, to make dependency on required libasr ready bebore plugin build,
masterport can define BUILD_DEPENDS additionally to RUN_DEPENDS for
opensmtpd option
! In D6607#139525, @fluffy wrote:
Instead of hardcoded mail/opensmtpd and dns/libasr it now configurable:
master port now have option to select between stable/devel opensmtpd,
and both opensmtpd ports have dependency via options between two
versions of libasr
Ah! Right you are. Can you change the description of it? It reads really strangely. Maybe "Depend on opensmtpd-devel (opensmtpd if off)" or something?
The problem here is that it shouldn't be necessary to install the master port for the slaves to work. That selection between opensmtpd and opensmtpd-devel should be present for the slaves as well.
This filter requires libasr so you'll have to leave one or the other in.
BTW, to make dependency on required libasr ready bebore plugin build,
masterport can define BUILD_DEPENDS additionally to RUN_DEPENDS for
opensmtpd option
Then opensmtpd should be a BUILD_DEPENDS as well.
I'd separate out the OPENSMTPD_DEVEL and SSL_PORT options and make them a part of everything, slave or master:
OPTIONS_DEFINE= OPENSMTPD_DEVEL SSL_PORT OPENSMTP_DEVEL_BUILD_DEPENDS_OFF= smtpctl:${PORTSDIR}/mail/opensmtpd OPENSMTP_DEVEL_RUN_DEPENDS_OFF= smtpctl:${PORTSDIR}/mail/opensmtpd OPENSMTP_DEVEL_BUILD_DEPENDS= smtpctl:${PORTSDIR}/mail/opensmtpd-devel OPENSMTP_DEVEL_RUN_DEPENDS= smtpctl:${PORTSDIR}/mail/opensmtpd-devel [SSL_PORT stuff] .if empty(SLAVE_PORT) [...]
And actually, by bringing in opensmtpd as a BUILD_DEPENDS, the SSL_PORT option should be unnecessary.
28.05.16 2:03, adamw (Adam Weinberger) пишет:
And actually, by bringing in opensmtpd as a BUILD_DEPENDS, the SSL_PORT
option should be unnecessary.
Yes, indeed, it will be available via opensmtpd[-devel] build dependency
It's getting difficult to track the changes here separate from D6084. Given that they are now getting intermingled, can you fold this into the other diff?
There is nothing wrong with the current indentation. Can you please leave it alone? It just makes the diff huge and hides the actual functional changes.
Moving PORTREVISION to slaves while keeping PORTVERSION in the master will no doubt end up in bumping PORTVERSION and forget to reset PORTREVISION.
Are we sure at this point that we'll always be willing to update them all together?
Can we please instead kill the master port and provide a Makefile.extras for the common stuff?
I'm stressing on this point cause it'll allow different people to efficiently maintain / test / update extra ports independently. I'm pretty sure none of us has a setup that allows him to properly test all extras.
Thoughts?
Right now they're all pulled from the same tarball. Unless you're proposing that each extra should have its own tarball and possibly be from different git revisions (which I'm totally fine with), I don't see how a Makefile.extras is any different from a master/slave.
- Pass master's CONFIGURE_ARGS to slaves
- Option to select stable/devel opensmtpd as build/run dependency (and corresponding libasr via this)
- Pass pkgconfig dependency only to supports which need it.
Yes, I suggest that every extra should be able to be updated independently, hence different distfiles. I know it's a waste of space and time, but the alternative is scarier...
I'm fine with the latest diff. IMHO we can postpone infrastructural changes to a later discussion / differential.
Thanks so much Dima and Adam for your time and effort!