Created attachment 434406 [details] net-misc/kea-1.0.0.ebuild Hi, please find attached a update to kea ebuild. This ebuild adds USE flags for mysql and postgresql. I had to modify the optimization CFLAGS since -O2 on GCC 5.3 causes the d2 server (kea-dhcp-ddns) to consume 100% of a core (see http://comments.gmane.org/gmane.network.dhcp.isc.kea.user/109 and http://kea.isc.org/ticket/4243). I tested and ran the server on ~amd64 with USE="mysql" , with host reservation and leases in mysql plus ddns updates to a bind server. I only compiled it with postgres USE flag, it builds on my setup but I can't guarantee it works as expected.
kea-1.0.0 is in the tree since a while. So I will close this bug report. Please reopen, if I misunderstood you. $ head /usr/portage/net-misc/kea/ChangeLog # ChangeLog for net-misc/kea # Copyright 1999-2016 Gentoo Foundation; Distributed under the GPL v2 # (auto-generated from git log) *kea-1.0.0 (07 Jan 2016) 07 Jan 2016; Tony Vroon <chainsaw@gentoo.org> +kea-1.0.0.ebuild, +metadata.xml: Adding 1.0.0 to tree.
Oh my mistake.. you added a new USE flag...
(In reply to Xavier Guerin from comment #0) > This ebuild adds USE flags for mysql and postgresql. I see a slightly different approach to that taken by Kim B. Sandalsen in bug #581976, where the dependency is on DBD-MySQL instead of on the MySQL client library virtual. Does it not use DBD? (If it does, both database dependencies should be the relevant DBD module) > I had to modify the optimization CFLAGS since -O2 on > GCC 5.3 causes the d2 server (kea-dhcp-ddns) to consume 100% of a core (see > http://comments.gmane.org/gmane.network.dhcp.isc.kea.user/109 and > http://kea.isc.org/ticket/4243). -O0 is very harsh on others who may not be using GCC 5.3 at all. I am not convinced this is a fair way to handle a compiler bug. > I tested and ran the server on ~amd64 with USE="mysql" , with host > reservation and leases in mysql plus ddns updates to a bind server. I only > compiled it with postgres USE flag, it builds on my setup but I can't > guarantee it works as expected. Thank you, a compile test is fair as long as it is declared.
Created attachment 442404 [details] ldd of main kea binaries ldd of main kea binaries, I only omitted shell scripts used for starting or stoping kea
Created attachment 442406 [details] equery files kea
Created attachment 442408 [details] new version of the ebuild with test on gcc version
(In reply to Tony Vroon from comment #3) > (In reply to Xavier Guerin from comment #0) > > This ebuild adds USE flags for mysql and postgresql. > > I see a slightly different approach to that taken by Kim B. Sandalsen in bug > #581976, where the dependency is on DBD-MySQL instead of on the MySQL client > library virtual. > Does it not use DBD? (If it does, both database dependencies should be the > relevant DBD module) I saw that too. When I look at what is linked in the binaries, it only uses the C client. I do not see perl files in kea, so I'd say it is not required. (btw I also have DBD-mysql installed as it is a dependency for mariadb compiled with use="perl") > > > I had to modify the optimization CFLAGS since -O2 on > > GCC 5.3 causes the d2 server (kea-dhcp-ddns) to consume 100% of a core (see > > http://comments.gmane.org/gmane.network.dhcp.isc.kea.user/109 and > > http://kea.isc.org/ticket/4243). > > -O0 is very harsh on others who may not be using GCC 5.3 at all. I am not > convinced this is a fair way to handle a compiler bug. I agree, I tested with GCC 5.4 and experienced the same issue. I modified the ebuild to check on GCC version and act accordingly (i.e. change the compile flag when we know it will fail with the default one). I might though miss issues with other versions, a warning after the installation could be useful here. > > > I tested and ran the server on ~amd64 with USE="mysql" , with host > > reservation and leases in mysql plus ddns updates to a bind server. I only > > compiled it with postgres USE flag, it builds on my setup but I can't > > guarantee it works as expected. > > Thank you, a compile test is fair as long as it is declared. I don't know if the ebuild will reach the tree for this version as the next one is scheduled for the end of August.
Should it be closed as current kea-1.2 exposes both flags?
Don't just put it in the tree, close the bug. Bad Tony.
