Let me know if I can help in anyway. It's a pity the package.yml is such a mess, it's not your fault I know, but if it could be cleaned up a bit that would be great!

Bonne chance!

This revision now requires changes to proceed.Nov 6 2017, 6:37 PM

In D7#20782, @joebonrichie wrote:

Needs to be rebased against R444:f1ee4191989f. Also testing, lots of testing of all reverse dependencies.

Let me know if I can help in anyway. It's a pity the package.yml is such a mess, it's not your fault I know, but if it could be cleaned up a bit that would be great!

Bonne chance!

Actually I am glad you ask because some help would indeed be more than welcome!
I did mess up to package.yml to add python 3 alongside with python 2 and I haven't found a cleaner way to do it.

I made a few quick checks against existing packages and it seems to be okay but with python 3 it is broken and the application that requires lot of resources to build.
If you find a way to make it working, it would be great then I'd be very happy to make deeper testing.

padraig added a subscriber: padraig.Nov 8 2017, 9:50 AM

I have some time to look at this now

boost

Actually rebase now (hopefully)

Looking into this patch from fedora
http://pkgs.fedoraproject.org/cgit/rpms/boost.git/tree/boost-1.63.0-dual-python-build-v2.patch

How is this looking in terms of revdep rebuilds? The current version is now blocking package updates (RStudio at this stage)

Bump to 1.66.0

This is based on the current version of the boost package (with python 2 support only)

In D7#33071, @kyrios123 wrote:

Funny you ask, I was just looking at this (I think our version of boost is getting old). Since I did not succeed building a package for both python2 & python3, I propose to first upgrade boost to 1.66.0 for python 2 only so at least we can go forward. Here is my status:

That is sensible to me, also we don't necessarily have to go all the way to 1.66.0 as software tends to lag behind a bit in supporting it, so being right up to date can be more of a burden than benefit. But if you can rebuild it all against it, then no worries.

Rstudio source/tags are gone so that can only be updated for a rebuild.

@sunnyflunk Your magic script did a wonderful job ?
I updated the still draft status on the above task. There aren't much issues, but there are a few.
I had to update a few applications and I found patches for some other (note that about 50% of the patches are for gcc 7, not for boost).

I still have to place the pre-requistes for mongodb and hugin in the list and sort out the rebuilds, but first I have to figure out why blender doesn't build anymore here are the logs.

In D7#33102, @kyrios123 wrote:

@sunnyflunk Your magic script did a wonderful job ?
I updated the still draft status on the above task. There aren't much issues, but there are a few.
I had to update a few applications and I found patches for some other (note that about 50% of the patches are for gcc 7, not for boost).
I still have to place the pre-requistes for mongodb and hugin in the list and sort out the rebuilds, but first I have to figure out why blender doesn't build anymore here are the logs.

Always ensure you do the validation checks as well (not doing that makes the automation worse imo), without checking you'll see things dropping freetype2 for example which is something you don't want to push (I even check the build logs of the .0 KDE updates to see dep changes as well). As packages get rebuilt and updated more frequently the task becomes much easier. My roadblock is building the packages xD

@sunnyflunk the full rebuild is over.

The only issues are libcutl and vcmi which fortunately don't have reverse dependencies. I need some help for these 2 and peercoin which I had to update to 0.6.2 RC3.

Could you also check on your local repository which packages actually use libboost-devel ? I made my list from the reverse dependencies but I see some packages having it in their builddeps that aren't in my list (i.e: Scribus)

i think component should be programming.library :)

In D7#33196, @livingsilver94 wrote:

i think component should be programming.library :)

Agreed ! I'll change this if/when the diff will be accepted

In D7#33195, @kyrios123 wrote:

Could you also check on your local repository which packages actually use libboost-devel ? I made my list from the reverse dependencies but I see some packages having it in their builddeps that aren't in my list (i.e: Scribus)

In theory these don't need rebuilding (if they don't depend on the package, magic abi_used_libs searching to check all are rebuilt is definitely necessary), but if they don't build, can cause an issue when they do need rebuilding.

[ ] abiword
[ ] akonadi-calendar
[ ] akonadi-contacts
[ ] akonadi-import-wizard
[ ] akonadi-mime
[ ] akonadi-notes
[ ] akonadi-search
[ ] akonadi
[ ] akregator
[ ] ardour
[ ] arx-libertatis
[ ] calendarsupport
[ ] cgal
[ ] chromaprint
[ ] codeblocks
[ ] compiz
[ ] digikam
[ ] et
[ ] eventviews
[ ] flightgear
[ ] fritzing
[ ] gnote
[ ] highlight
[ ] incidenceeditor
[ ] inkscape
[ ] kactivities-stats
[ ] kactivities
[ ] kactivitymanagerd
[ ] kaddressbook
[ ] kalarmcal
[ ] kdepim-apps-libs
[ ] kdepim-runtime
[ ] kdevelop
[ ] kget
[ ] kleopatra
[ ] kmail-account-wizard
[ ] kmail
[ ] kmailtransport
[ ] kontact
[ ] kontactinterface
[ ] korganizer
[ ] libabw
[ ] libcdr
[ ] libe-book
[ ] libetonyek
[ ] libexempi
[ ] libfreehand
[ ] libftdi1
[ ] libgltf
[ ] libkdepim
[ ] libkgapi
[ ] libkleo
[ ] libktorrent
[ ] libmspub
[ ] libmwaw
[ ] libodfgen
[ ] libpagemaker
[ ] libqxp
[ ] librecad
[ ] librevenge
[ ] libvisio
[ ] libvsqlitexx
[ ] libwpd
[ ] libwps
[ ] libzmf
[ ] mailcommon
[ ] mailimporter
[ ] mdds
[ ] messagelib
[ ] mpd
[ ] mumble
[ ] nemiver
[ ] nixnote2
[ ] openxcom
[ ] otclient
[ ] pim-data-exporter
[ ] pimcommon
[ ] plasma-desktop
[ ] polybar
[ ] qbittorrent
[ ] scribus
[ ] sigil
[ ] simgear
[ ] supertux
[ ] wiznote
[ ] xboxdrv
[ ] xsd
[ ] yaml-cpp

Try building them from git master

In D7#33344, @sunnyflunk wrote:

Try building them from git master

I tried to build libcutl from git but it fails at the bootstrap stage and I'm too ignorant to find out what is missing. I suspect, it's only used for xsd. If I'm right, we could also use the bundled version until libcutil 1.11.0 is released.
vcmi build goes a little further from git, but I still it fails.
compizfailing to build is not caused by boost since I don't succeed rebuilding it at all (log file).

In D7#33596, @kyrios123 wrote:

I tried to build libcutl from git but it fails at the bootstrap stage and I'm too ignorant to find out what is missing. I suspect, it's only used for xsd. If I'm right, we could also use the bundled version until libcutil 1.11.0 is released.

We can probably drop libcutl, provided a workable xsd can be built. You may need to not push the latest boost version if vcmi can't be resolved

libcutl and xsd are no longer relevant to this as I'm deprecating libcutl and rebuilt xsd without as it will only cause further issues in future.

Back to 1.65.1 since vcmi is a blocker for 1.66.0

@sunnyflunk I did rebuild this list against boost 1.65.1. There are a few packages that I don't need to update to a newer version or to patch anymore. Everything is going smoothly since I already learned from my previous build attempt.
If you accept this patch, I can publish everything tomorrow so there are enough time to test on unstable before the next sync.

padraig mentioned this in D2202: Enable Python 3 support for networkx..Mar 2 2018, 9:54 AM

So how is this looking in terms of conflicting with the wxwidgets testing? A quick glance only seems like poedit as being in both lists. I would really like to get this in and rstudio updated.

In D7#37451, @sunnyflunk wrote:

So how is this looking in terms of conflicting with the wxwidgets testing? A quick glance only seems like poedit as being in both lists. I would really like to get this in and rstudio updated.

Actually I was waiting for someone to approve the task but since there was no reaction and wxwidgets is shorter and has a smaller impact, I asked ikey if I can start with this one.
The following apps are in both lists and I know I'll have to rebuild at least libreoffice since I updated it in the meantime: