both mariadb-scripts-5.3.7 and mariadb-server-5.3.7 install /usr/local/bin/mysqld_multi with different path entries. mariadb-scripts: /usr/local/var[1-6] mariadb-server: /var/db/mysql[1-6] This leads to checksum error (pkg_info -g) in the previously installed port.
Maintainer of databases/mariadb-scripts, Please note that PR ports/171338 has just been submitted. If it contains a patch for an upgrade, an enhancement or a bug fix you agree on, reply to this email stating that you approve the patch and a committer will take care of it. The full text of the PR can be found at: http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/171338 -- Edwin Groothuis via the GNATS Auto Assign Tool edwin@FreeBSD.org
State Changed From-To: open->feedback Awaiting maintainers feedback (via the GNATS Auto Assign Tool)
Still a plist problem... will try to fix it this week. -- *geoffroy desvernay* C.R.I - Administration systèmes et réseaux Ecole Centrale de Marseille Tel: (+33|0)4 91 05 45 24 Fax: (+33|0)4 91 05 45 98 dgeo@centrale-marseille.fr
This is a multi-part message in MIME format.
State Changed From-To: feedback->open Maintainer has fixed.
Could someone commit this SVP ? -- *geoffroy desvernay* C.R.I - Administration systèmes et réseaux Ecole Centrale de Marseille Tel: (+33|0)4 91 05 45 24 Fax: (+33|0)4 91 05 44 26
Responsible Changed From-To: freebsd-ports-bugs->feld I'll take it.
Anyone to take a look at this ?
Created attachment 146735 [details] patch against databases/mariadb-scripts Last patch attached in GNATS
Last I saw the patch didn't solve the problem completely. There are inappropriate things being installed by the different mariadb ports. They should perfectly model the mysql (5.1 ?) ports, but they do not. I was working on it but mariadb was refusing to build and produce identical output when given the exact same build flags as the mysql port version it is sister to, and then I had given up due to $work projects. I'll take a look again at the patch you provided and see if it is sufficient to make them installable for now, I guess. Users should really be encouraged to move to a newer MariaDB release :-)
Ok, I tried looking at this again and here's what I ran into: I applied your patch, but cannot build mariadb-scripts because I can't build mariadb-server. mariadb-server has plist problems. Fixing the plist problems gets you one step further, but then mariadb-server conflicts with mariadb-client on include/mysql/client_plugin.h. If I remove that from mariadb-server, mariadb-client won't build successfully because of the way the client port is a slave of the server port. It's a big mess. All three ports need to be fixed to solve this problem. I literally cannot test your mariadb-scripts patch with poudriere testport until the other issues with mariadb-server and mariadb-client are solved. Sorry for the bad news :-(
Well, here we don't use mariadb 5.3 anymore and I don't have much time to look into this. Maybe removing these ports would be the less worse thing to do for now, maybe with lines in MOVED: upgrade to 5.5 should not break (unless using PBXT maybe, but it wasn't enabled by default). I'm not aware of how to delete ports... svn diff is sufficient to submit here ? Should I change maintainer to ports@ and DEPRECATE them ? Thank you for your help, and sorry to not be able to correct these for now
Created attachment 148424 [details] Build log mariadb-client 5.3.12_5
Created attachment 148425 [details] Build log mariadb-server 5.3.12_5
A. SLAVE PORTS The mysql and mariadb ports split in a -client and a -server part but mysql51-* and mariadb-* (5.3) split in 3 parts: -client, -server and -scripts. MariaDB 5.3 is a fork of MySQL 5.1 so it should align with the MySQL 5.1 port. B. DEPENDENCIES The newer version ports (5.5, 5.6) include the scripts into the -server port. There's an issue altogether with this as none of the ports containing the scripts pull in sufficient deps to actually run the scripts. All of perl, p5-DBD and p5-DBI-mysql(51) must be available to run the perl scripts of the distribution. mysql51-scripts pulls in perl via the mysql51-scripts/Makefile but pulls in p5-DBD-mysql51 via the mysql51-server/Makefile Neither mariadb-server MARIADB_SCRIPTS_SLAVE nor mariadb-client depend on p5-DBD-mysql so this can be seen as broken to begin with. C. BUILD FAILURE I was able to build both mariadb-client and -server 5.3.12_5 without problems on 10.1-RC2, buildlogs attached
Our poudriere also build mariadb-{client,scripts,server} without problem here (on 8.4 amd64 at least) - current port tree (r371168) http://ftp.ec-m.fr/pkg/logs/bulk/8amd64-default-bdd/2014-10-19_11h53m21s/logs/mariadb53-client-5.3.12_5.log http://ftp.ec-m.fr/pkg/logs/bulk/8amd64-default-bdd/2014-10-19_11h53m21s/logs/mariadb53-scripts-5.3.12_5.log http://ftp.ec-m.fr/pkg/logs/bulk/8amd64-default-bdd/2014-10-19_11h53m21s/logs/mariadb53-server-5.3.12_5.log
Created attachment 148473 [details] Build logs of mariadb-{client,server,scripts}-5.3.12_5 Replaced the earlier buildlogs. All in one now
Created attachment 148474 [details] Patch for mariadb-server Patch removes duplicates between -scripts and -server, copying as much as possible the mysql51-server and -client examples.
Comment on attachment 148474 [details] Patch for mariadb-server maintainer-tested-and-approved :)
(In reply to geoffroy desvernay from comment #20) Patch does not apply cleanly. root@skeletor:/usr/ports # fetch -o /tmp/mariadb.diff "https://bz-attachments.freebsd.org/attachment.cgi?id=148474" fetch: https://bz-attachments.freebsd.org/attachment.cgi?id=148474: size of remote file is not known /tmp/mariadb.diff 1279 B 16 MBps 00m00s root@skeletor:/usr/ports # svn patch /tmp/mariadb.diff U databases/mariadb-server/Makefile > applied hunk @@ -85,10 +85,6 @@ with offset 3 C databases/mariadb-server/pkg-plist > rejected hunk @@ -395,3 +385,4 @@ Summary of conflicts: Text conflicts: 1
Created attachment 152516 [details] mariadb53-client build failure mariadb53-client fails to build on FreeBSD 10.0 via poudriere
Created attachment 152604 [details] updated patch This one applies well. I a maintainer change to Bernard (as discussed with him some time ago)
Just tried a poudriere bulk for both 101amd64 and 100amd64. The 10.0 build fails on the my_gethostbyname_r macro. Build output 10.0: checking style of gethost* routines... checking whether madvise is declared... yes other checking style of gethostbyname_r routines... other checking 3 argument to gethostbyname_r routines... char 10.1: checking style of gethost* routines... checking whether madvise is declared... yes other checking style of gethostbyname_r routines... glibc2 checking 3 argument to gethostbyname_r routines... char In 10.0 leading to ../include/my_net.h:113:74: note: expanded from macro 'my_gethostbyname_r' #define my_gethostbyname_r(A,B,C,D,E) gethostbyname_r((A),(B),(C),(D),(E)) Looks like glibc1 vs glibc2 style gethostbyname_r Not quite sure yet why it is behaving like this, further digging is required
see https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=197679 for an interim solution
A commit references this bug: Author: feld Date: Thu Mar 26 19:24:18 UTC 2015 New revision: 382337 URL: https://svnweb.freebsd.org/changeset/ports/382337 Log: Fix building on FreeBSD 10+ Fix conflicts between client, server, and script packages Shebangfix The plists now are closer aligned with the mysql51-* ports PR: 197679 PR: 171338 Changes: head/databases/mariadb-client/Makefile head/databases/mariadb-client/pkg-plist head/databases/mariadb-server/Makefile head/databases/mariadb-server/pkg-plist