build: fix symlinks in lv2dir

[puleglot]

Fixes loading of lv2 plugin and its gui.

[falkTX]

Which ticket/issue does this fix?
It has been working so far as-is for many people, I think applying these changes will break existing setups.

[puleglot]

This should fix regression introduced by d18a873 and 75f33e0

[puleglot]

Downstream bug: https://bugs.gentoo.org/779343

[gitterdude]

The last working git version in Ubuntu 18.04 right now is 4e69944; just prior to d18a873
The newer ones fail as stated in the gentoo report.

[falkTX]

So seems fine to me, a regression from a recent commit.
I added a little comment about something that needs attention, rest seems fine.
Thanks

[boomshop]

Great, thanks a bunch for taking care of this one @falkTX and @puleglot

Please let me know when this fix is in a state able to be merged and I'll pull the trigger on it.

[falkTX]

The changes look good
Though I did not verify if the build is indeed fixed (would be great to have some CI/automated testing..)

[nick87720z]

Fixes #294

I tried to edit symlink to point to libcalf.so. This made it loadable in carla, but without special gui.
PulseEffects keeps telling that these plugins (incl bass_enhancer) was not found.

Update: I can confirm, that commit from PR fixes both plugin loading and custom gui, at least in carla.
But pulseeffects still cries.
Tested puleglot's repo/branch/commit with live ebuild in gentoo:

EVCS_OFFLINE= \
EGIT_OVERRIDE_REPO_CALF_STUDIO_GEAR_CALF=https://github.com/puleglot/calf \
EGIT_OVERRIDE_BRANCH_CALF_STUDIO_GEAR_CALF=fix-symlinks \
EGIT_OVERRIDE_COMMIT_CALF_STUDIO_GEAR_CALF=078f24797183acc21d49730b33fa96c103f4aaa4 \
emerge -av1 --nodeps --quiet-build=n calf

[puleglot]

But pulseeffects still cries.

What problem with pulseeffects? I tested with Ardour only, but in theory calf should be loadable by any other LV2 host.

[zettberlin]

I tried to edit symlink to point to libcalf.so. This made it loadable in carla, but without special gui.

Same here.
I created a symlink from libcalf.so to calf.so. Plugins load and can be edited with a generic host UI but their own GUI does not load. No entry in Carlas log and none in Ardours log either...

[zettberlin]

Tried again with this branch, to no avail...
Filenames are as wrong as before in /usr/local/lib/calf (no calf.so but libcalf.so) and Carla and Ardour complain.

What do I do wrong?

[zettberlin]

I now set a symlink from python to python3, it does look a little different in Carla, as in, now I get something in the logs:

Will use UI-Bridge for 'GxAmplifier-Stereo-X', binary: "/usr/lib/carla/carla-bridge-lv2-x11"
ALL OK!
Will use UI-Bridge for 'Calf Flanger', binary: "/usr/lib/carla/carla-bridge-lv2-gtk2"
ALL OK!
Carla assertion failure: "fLibFilename.isNotEmpty()" in file CarlaBridgeFormat.cpp, line 104
Failed to load UI binary, error was:
(null)

As you see: the Guitarix GUI loads OK, Calf Flanger does not...

[kisaragi-hiu]

Arch Linux, but this should also replicate everywhere else:

• git clone https://github.com/calf-studio-gear/calf calf
• cd calf
• Remove the existing Calf installation, if any
• ./autogen.sh
• make
• sudo make install
• Open Ardour and insert Calf Compressor
• Observe the error --- Ardour's log only shows one line, that it failed to instantiate
• Open REAPER, for that matter, and insert Calf Compressor
• Observe even the generic UI fails to load

This patch seems to be necessary to get master builds to load again.

---

Filenames are as wrong as before in /usr/local/lib/calf (no calf.so but libcalf.so)

.../lib/calf/libcalf.so is the correct name. .../lib/lv2/calf.lv2/calf.so being a symlink to inexistant .../lib/calf/calf.so is the incorrect part.

I tried to edit symlink to point to libcalf.so
I created a symlink from libcalf.so to calf.so

Creating symlinks for libraries yourself is generally not a reliable way to get things to load properly.

---

Would it help if GitHub Actions is set up to at least check that ./autogen.sh; make succeeds?

[JohannesLorenz]

As requested, I was able to run this PR in CI (on a temporary branch) with success for linux-gcc, linux-clang, macos and mingw64: https://github.com/calf-studio-gear/calf/actions/runs/6578998596 .

If I understand the comments right, this is safe to merge now?

[JohannesLorenz]

Friendly reminder that I have an open question in the above post. Would be nice if anyone could help here.

[kisaragi-hiu]

Thank you for the maintenance; I'll take a look tomorrow to see how Calf builds on my machine (now using KDE Neon) before and after this patch.

[kisaragi-hiu]

I've tested again and the results still agree with my previous result, where the plugin does not load without this PR, and does with it applied.

With current master (4c0d805):

git clone https://github.com/calf-studio-gear/calf calf
cd calf

./autogen.sh
./configure
make # assuming lv2 and other dependencies are already present
sudo make install

Then open a DAW and try to load a Calf plugin. It fails to load.

With this PR applied:

# cleanup
sudo make uninstall
git clean -Xdf
# apply this PR onto the just-cloned copy
gh pr checkout 290 # using github cli
git rebase origin/master # just fast forward for the test

./autogen.sh
./configure
make # assuming lv2 and other dependencies are already present
sudo make install

Now open eg. Ardour and try to load a Calf plugin. It loads successfully.

As an aside, distro-provided Calf packages don't (yet) hit this issue because they generally use the latest tagged release, which is 41a2b7f (0.90.3) before the commits introducing the regression identified by OP.

[JohannesLorenz]

@kisaragi-hiu Thanks a lot! I guess we can just merge this PR then?

[JohannesLorenz]

Will merge in 3 days if no one will disagree.