[NEW PORTS] databases/mariadb101-{server,client}: 10.1.10

[brnrd]

Summary

Proposed commit log:

[NEW PORTS] databases/mariadb101-{server,client}: 10.1.10

MariaDB is one of the most popular database servers in the world. It’s made
by the original developers of MySQL and guaranteed to stay open source.
Notable users include Wikipedia, Facebook and Google.

MariaDB turns data into structured information in a wide array of
applications, ranging from banking to websites. It is an enhanced, drop-in
replacement for MySQL. MariaDB is used because it is fast, scalable and
robust, with a rich ecosystem of storage engines, plugins and many other
tools make it very versatile for a wide variety of use cases.

MariaDB is developed as open source software and as a relational database
it provides an SQL interface for accessing data. This latest versions of
MariaDB also include GIS and JSON features.

WWW: http://mariadb.org/

Reviewed_by:	koobs (mentor), feld (mentor)
Approved by:	
Differential_Revision:	D3953

Test Plan

• portlint -AC (clean)
• poudriere testport -client 10.2amd64 -server 10.2amd64 -client 9.3amd64 -server 9.3amd64
• Build with various options (pending)

[brnrd]

This requires the removal of the MariaDB 5.3 ports before the new ports can be added.

[antoine]

Please create seperate reviews for mariabd* addition and USES=mysql, those changes are orthogonal

[brnrd]

Separate new ports from Uses refactor

[brnrd]

Add diffs for the correct MariaDB version

[brnrd]

In D3953#87513, @antoine wrote:

Please create seperate reviews for mariabd* addition and USES=mysql, those changes are orthogonal

See D4166

[brnrd]

Update patches to MariaDB 10.1.9

[brnrd]

Update MariaDB 10.1 port

• Update to latest version 10.1.10
• Add upstream patch for json.cpp clang failure
• Use regular strip target

[brnrd]

databases/mariadb101-client/Makefile

16

That's the intention. Adding portname[0-9]* for all ports as this quenches portlint warnings.

databases/mariadb101-client/files/patch-cmake_jemalloc.cmake

6–22

Not at the moment. jemalloc is in base in FreeBSD and the MariaDB doesn't detect that correctly

[brnrd]

Fixes for the remaining comments

• Use variables to construct INSTALL_CONFLICTS
• Work around missing install/strip target in -client

[brnrd]

Refactor and upstream jemalloc patch

[brnrd]

databases/mariadb101-client/files/patch-cmake_jemalloc.cmake

4–20

Refactored patch and submitted it upstream

[brnrd]

[[ https://mariadb.atlassian.net/browse/MDEV-9350 | [PATCH] Fix jemalloc detection for FreeBSD ]]

[koobs]

Hmm, am not a fan of negative options, in this case WITHOUT_SSL.

OPTIONS_RADIO allows for none, one or more options to be selected (contrary to what it's name might imply) and having grepped the ports tree for other references to WITHOUT_SSL, none are used in the context of OPTIONS in the same way, other than other mysql/mariadb ports.

Other ports solve this as follows:

./ftp/vsftpd-ext/Makefile:.if ${PORT_OPTIONS:MVSFTPD_SSL} && !defined(WITHOUT_SSL)

Unfortunately this also precludes the use of OPTIONS helpers, which isn't by itself a warrant for a WITHOUT_SSL option.

Having said that because the other mysql/mariadb ports already use this construct, im happy for it to proceed for now, but I would encourage you to refactor these ports to remove the need for WITHOUT_SSL as an option, while retaining support for users turning SSL support off via WITHOUT_SSL

[brnrd]

Fix OpenSSL and TokuDB options

• Upstream has removed building without SSL
• TokuDB wasn't building at all before
• Make more storage modules configurable

[brnrd]

Replaced the proposed commit message, the one that was there was no longer accurate. Replaced with http://mariadb.org/about/ and replaced The latest with This latest.

svn rm-ed the whole thing, then repo-copied mariadb100- dirs and fixed remaining renames in files/patch*. Just ran both though poudriere testport and they are OK still.

[koobs]

• All patches need what/why breadcrumbs (comments) with upstream references/links if available (mailing lists, issues, commits)
• We (project, not royal) strongly prefer not to use bundled libraries. I would prefer to see WOLFSSL as an OPTION{_DEFAULT} with SSL (OPEN) as an alternative option, in an OPTIONS_SINGLE group

[brnrd]

Annotate patches, don't use bundled libs

[brnrd]

Annotate patches, don't use bundled libs, work around arc failure

[brnrd]

To get the patch in using arc after repocopy I had to create 2 empty patches
databases/mariadb101-server/files/patch-extra_yassl_taocrypt_src_integer.cpp
databases/mariadb101-server/files/patch-scripts_CMakeLists.txt

[koobs]

databases/mariadb101-client/files/patch-scripts_mysql_config.sh

7–9 ↗

(On Diff #12329)

Can't these be passed into CMAKE_* ?

What is the root cause for incorrect/missing in the first place?

[brnrd]

Update patch for MariaDB 10.1.11

• Patch for CyaSSL no longer required
• Patch for json included upstream
• Add gssapi auth plugin (with upstream fix)

[brnrd]

databases/mariadb101-client/files/patch-scripts_mysql_config.sh

7–9 ↗

(On Diff #12329)

$pkglibdir is dir to MariaDB's libs (default /usr/local/lib/mysql)
LIBS_FOR_CLIENTS is the libraries themselves, so I can't see a way to fix this.
There's an error here though: /usr/local -> %%LOCALBASE%%

[koobs]

databases/mariadb101-client/files/patch-scripts_mysql_config.sh

7 ↗

(On Diff #13104)

Was changing this from:

+libs="$libs -L%%LOCALBASE%%/lib"
+libs_r="$libs_r -L%%LOCALBASE%%/lib"
+embedded_libs="$embedded_libs -L%%LOCALBASE%%/lib"

intentional?

databases/mariadb101-server/Makefile

32

No default?

[brnrd]

Fix LOCALBASE hard coding

[brnrd]

databases/mariadb101-server/Makefile

32

Added OPTIONS_DEFAULT= GSSAPI_BASE

[brnrd]

databases/mariadb101-client/files/patch-scripts_mysql_config.sh

12 ↗

(On Diff #13261)

Fixed

[brnrd]

Set a default for GSSAPI libs

[koobs]

Pending review by @feld, which I recommend given at least the size of this port.

I note that the replacement of hardcoded /usr/local paths in patches will be deferred to post-initial commit.

[feld]

Builds for me, approved