Update contrib/subversion to the latest LTS version (1.10.6)

[gbe]

Summary

Update subversion to its latest LTS version, which is 1.10.6.
apr was also updated to 1.7.0 and apr-util to 1.6.1. The FreeBSD specific additions
to util.c where backported.

Test Plan

make tinderbox and an amd64 installation with checkouts of head, ports and docs including verification, that $FreeBSD$ Ids are correctly expanded.

[lwhsu]

Is there a suggested check/test list for updating the svn(lite) in base?

[gbe]

Not as far as I know. I tested a "make buildworld" on amd64 and a local invocation of the freshly compiled svnlite, which worked. I also tested a crossbuild for arm64.

In D23383#515771, @lwhsu wrote:

Is there a suggested check/test list for updating the svn(lite) in base?

[lwhsu]

In D23383#515795, @gbergling_gmail.com wrote:

Not as far as I know. I tested a "make buildworld" on amd64 and a local invocation of the freshly compiled svnlite, which worked. I also tested a crossbuild for arm64.

In D23383#515771, @lwhsu wrote:

Is there a suggested check/test list for updating the svn(lite) in base?

Ok this reminds me that I think this kinds of change needs a make tinderbox build to check all the architecture are fine.
Also, the main task of svnlite is for getting the latest freebsd source code (and ports and docs), could you do some checkout and update tests? And verify the things like $FreeBSD$ tag is expanded correctly?

Thanks for working on this!

[gbe]

In D23383#515834, @lwhsu wrote:

In D23383#515795, @gbergling_gmail.com wrote:

Not as far as I know. I tested a "make buildworld" on amd64 and a local invocation of the freshly compiled svnlite, which worked. I also tested a crossbuild for arm64.

In D23383#515771, @lwhsu wrote:

Is there a suggested check/test list for updating the svn(lite) in base?

Ok this reminds me that I think this kinds of change needs a make tinderbox build to check all the architecture are fine.
Also, the main task of svnlite is for getting the latest freebsd source code (and ports and docs), could you do some checkout and update tests? And verify the things like $FreeBSD$ tag is expanded correctly?

Thanks for working on this!

I updated the testplan and start the build this evening. I post the results, when they are ready.

[gbe]

So, I just finished the testplan. A freshly -CURRENT from yesterday with the D23383 applied resulted in the following svnlite and the checkouts of head, ports and docs where successful. The tinderbox build was failing due to an "arm LINT-V7 kernel failed" error, but the "make universe" target was successful. I would think that the testplan was passed.

gbergling@fbsd-dev:~ % svnlite --version
svn, version 1.10.6 (r1863367)
   compiled Feb  6 2020, 21:06:09 on bikeshed-rgb-freebsd

Copyright (C) 2019 The Apache Software Foundation.
This software consists of contributions made by many people;
see the NOTICE file for more information.
Subversion is open source software, see http://subversion.apache.org/

[lwhsu]

In D23383#516976, @gbergling_gmail.com wrote:

So, I just finished the testplan. A freshly -CURRENT from yesterday with the D23383 applied resulted in the following svnlite and the checkouts of head, ports and docs where successful. The tinderbox build was failing due to an "arm LINT-V7 kernel failed" error, but the "make universe" target was successful. I would think that the testplan was passed.

Thanks. As described in src/Makefile, the universe target will not error out when there is error immediately. I don't think the armv7 LINT failure is caused by svn update, but please check if there is any other bug in that. This is another thing, though. But please be sure to check all the buildworld targets fine on all architectures. If you could check running this on some non-x86 arches, it will be great.

For merging this to the tree, it would take some extra work at the committer side. We don't directly import contributed software to the contrib/ director, but put to vendor branch first and then merge, as https://www.freebsd.org/doc/en_US.ISO8859-1/books/developers-handbook/policies-contributed.html describes. It takes some time to take care these things. If you happen to have more time, please help me check if there is anything we should do with the local modification. Thank you.

[gbe]

In D23383#516982, @lwhsu wrote:

In D23383#516976, @gbergling_gmail.com wrote:

So, I just finished the testplan. A freshly -CURRENT from yesterday with the D23383 applied resulted in the following svnlite and the checkouts of head, ports and docs where successful. The tinderbox build was failing due to an "arm LINT-V7 kernel failed" error, but the "make universe" target was successful. I would think that the testplan was passed.

Thanks. As described in src/Makefile, the universe target will not error out when there is error immediately. I don't think the armv7 LINT failure is caused by svn update, but please check if there is any other bug in that. This is another thing, though. But please be sure to check all the buildworld targets fine on all architectures. If you could check running this on some non-x86 arches, it will be great.

For merging this to the tree, it would take some extra work at the committer side. We don't directly import contributed software to the contrib/ director, but put to vendor branch first and then merge, as https://www.freebsd.org/doc/en_US.ISO8859-1/books/developers-handbook/policies-contributed.html describes. It takes some time to take care these things. If you happen to have more time, please help me check if there is anything we should do with the local modification. Thank you.

The make universe was completely successful. I just checked the build-log.

--------------------------------------------------------------
>>> make universe completed on Thu Feb  6 09:13:14 CET 2020
                      (started Wed Feb  5 15:40:40 CET 2020)
--------------------------------------------------------------

I currently only have amd64 hardware, so I can't really test a non-x86 architecture. :( The vendor import could be a little tricky, because there are a lot of files that are not going to be imported, subversion tests and various non-unix support files (beos, win32, etc.). I didn't remember exactly which ones. Backports where only needed for util.c, but in r357299 and r357301 cem@ did some modifications to contrib/apr which should be reapplied or better send upstream.

If I can provide any more help, just let me know.

[gbe]

Any update on this? I would think that the subversion sources could be integrated from this diff into the vendor repository. So it would be possible to go the specified way which is described in the developers handbook.

[gbe]

Superseded with the import of Subversion 1.14.0 LTS.