Seems that it doesn’t break the app on Big Sur. I think that’s enough.

Has anyone checked to see if Github offer an earlier MacOS builder?

It's probably worth always running the minimum and latest versions of a platform we intend to say we support. MacOS, of course, being the right pain in terms of constant change.

Does this also need changes to codeql-analysis.yml? That is still referring to Qt 5.15.2 for Mac.

Just got a positive feedback from the tester. Seems to work. Not sure if we can use an even earlier version of Qt, but that should be possible.

Should we try qt 5.9 (but for me it seems to be quite old). It might even support macOS 10.10

Does this also need changes to codeql-analysis.yml

Not necessarily.

Should we try qt 5.9

Do we know which version of Qt corrados used to compile the binaries?

Do we know which version of Qt corrados used to compile the binaries?

On my PC, the Qt DLLs bundled with Jamulus 3.6.2 are version 5.15.1. That works for me on Win7.

On my Mac, looking in /Applications/Jamulus.app/Contents/Frameworks/QtCore.framework/Versions/5/Resources/Info.plist I can see that it bundled version 5.9.4

Ok. So we could probly also use qt 5.9. What about 5.9.9 does it work?

Ok. So we could probly also use qt 5.9. What about 5.9.9 does it work?

Probably, as it's just a minor version number change. I could try it tomorrow.

Ok. Great!

OK, I tried 5.9.9 here pljones#7 and it failed.... but for the wrong reason. The MacOS build ended with:

ld: warning: directory not found for option '-F/Users/qt/work/install/lib'

and still claimed to have worked. I'm trying a re-run now...

OK, I tried 5.9.9 here pljones#7 and it failed.... but for the wrong reason. The MacOS build ended with:

ld: warning: directory not found for option '-F/Users/qt/work/install/lib'

and still claimed to have worked. I'm trying a re-run now...

Just been watching it, and that was just a warning, that an additional directory it was told to look for libs when linking wasn't found. So long as it still found all the libs it wanted to link, it shouldn't matter. If it couldn't, it would have been a hard error.

It looks like the warning is because it said /Users/qt/... when everything else is under /Users/runner/.... Not sure whether that's a bug in our build script or the Github environment.

I've just changed 5.12.9 to 5.9.9 and tried the build from scratch on my Mac.

Worked fine, and installed and ran successfully too. The Info.plist files correctly identified the bundled Qt as 5.9.9

Just as an interesting data point, MuseScore, which uses Qt, advertises itself as compatible with macOS 10.10 (Yosemite) onwards and Windows 7 onwards. The Mac build uses Qt 5.9.8 (although I'm sure 5.9.9 would be equally compatible, as a minor version change only), and the PC build uses Qt 5.9.9

I've just closely examined the build log on my Mac, and it shows exactly the same error message, so it's not a problem with Github or the action. This is the link step:

/Library/Developer/CommandLineTools/usr/bin/clang++ -stdlib=libc++ -headerpad_max_install_names  -arch x86_64 -Wl,-syslibroot,/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Develop
er/SDKs/MacOSX10.15.sdk -mmacosx-version-min=10.10 -Wl,-rpath,@executable_path/Frameworks -Wl,-rpath,/usr/local/opt/qt/5.9.9/clang_64/lib -o Jamulus.app/Contents/MacOS/Jamulus sound.o buffer.o channel.o
 client.o main.o protocol.o jamcontroller.o server.o serverlist.o serverlogging.o settings.o signalhandler.o socket.o soundbase.o util.o jamrecorder.o creaperproject.o cwavestream.o audiomixerboard.o ch
atdlg.o clientsettingsdlg.o connectdlg.o clientdlg.o serverdlg.o multicolorled.o levelmeter.o analyzerconsole.o bands.o celt.o celt_decoder.o celt_encoder.o celt_lpc.o cwrs.o entcode.o entdec.o entenc.o
 kiss_fft.o laplace.o mathops.o mdct.o modes.o pitch.o quant_bands.o rate.o vq.o A2NLSF.o ana_filt_bank_1.o biquad_alt.o bwexpander.o bwexpander_32.o check_control_input.o CNG.o code_signs.o control_aud
io_bandwidth.o control_codec.o control_SNR.o debug.o decoder_set_fs.o decode_core.o decode_frame.o decode_indices.o decode_parameters.o decode_pitch.o decode_pulses.o dec_API.o encode_indices.o encode_p
ulses.o enc_API.o apply_sine_window_FLP.o autocorrelation_FLP.o burg_modified_FLP.o bwexpander_FLP.o corrMatrix_FLP.o encode_frame_FLP.o energy_FLP.o find_LPC_FLP.o find_LTP_FLP.o find_pitch_lags_FLP.o
find_pred_coefs_FLP.o inner_product_FLP.o k2a_FLP.o LPC_analysis_filter_FLP.o LTP_analysis_filter_FLP.o LTP_scale_ctrl_FLP.o noise_shape_analysis_FLP.o pitch_analysis_core_FLP.o process_gains_FLP.o resi
dual_energy_FLP.o scale_copy_vector_FLP.o scale_vector_FLP.o schur_FLP.o sort_FLP.o warped_autocorrelation_FLP.o wrappers_FLP.o gain_quant.o HP_variable_cutoff.o init_decoder.o init_encoder.o inner_prod
_aligned.o interpolate.o lin2log.o log2lin.o LPC_analysis_filter.o LPC_fit.o LPC_inv_pred_gain.o LP_variable_cutoff.o NLSF2A.o NLSF_decode.o NLSF_del_dec_quant.o NLSF_encode.o NLSF_stabilize.o NLSF_unpa
ck.o NLSF_VQ.o NLSF_VQ_weights_laroia.o NSQ.o NSQ_del_dec.o pitch_est_tables.o PLC.o process_NLSFs.o quant_LTP_gains.o resampler.o resampler_down2.o resampler_down2_3.o resampler_private_AR2.o resampler
_private_down_FIR.o resampler_private_IIR_FIR.o resampler_private_up2_HQ.o resampler_rom.o shell_coder.o sigm_Q15.o sort.o stereo_decode_pred.o stereo_encode_pred.o stereo_find_predictor.o stereo_LR_to_
MS.o stereo_MS_to_LR.o stereo_quant_pred.o sum_sqr_shift.o tables_gain.o tables_LTP.o tables_NLSF_CB_NB_MB.o tables_NLSF_CB_WB.o tables_other.o tables_pitch_lag.o tables_pulses_per_block.o table_LSF_cos
.o VAD.o VQ_WMat_EC.o analysis.o mlp.o mlp_data.o opus.o opus_decoder.o opus_encoder.o repacketizer.o activity.o qrc_resources.o moc_channel.o moc_client.o moc_protocol.o moc_jamcontroller.o moc_server.
o moc_serverlist.o moc_settings.o moc_socket.o moc_soundbase.o moc_testbench.o moc_util.o moc_jamrecorder.o moc_creaperproject.o moc_signalhandler.o moc_audiomixerboard.o moc_chatdlg.o moc_clientsetting
sdlg.o moc_connectdlg.o moc_clientdlg.o moc_serverdlg.o moc_levelmeter.o moc_analyzerconsole.o   -F/usr/local/opt/qt/5.9.9/clang_64/lib -framework CoreFoundation -framework CoreServices -framework CoreA
udio -framework CoreMIDI -framework AudioToolbox -framework AudioUnit -framework Foundation -framework QtWidgets -framework QtGui -framework QtCore -framework DiskArbitration -framework IOKit -framework
 QtMacExtras -F/Users/qt/work/install/lib -framework QtNetwork -framework QtXml -framework QtConcurrent -framework OpenGL -framework AGL

ld: warning: directory not found for option '-F/Users/qt/work/install/lib'

Just investigating...

It appears to be something hard-coded within the homebrew installation of Qt. I would just ignore it, it's benign.

hard-coded within the homebrew installation of Qt

Are you really using homebrew here? I thought (at least the GH Action) used aqt to install Qt?

hard-coded within the homebrew installation of Qt

Are you really using homebrew here? I thought (at least the GH Action) used aqt to install Qt?

I think aqt must be using it behind the scenes, as I found the references in /usr/local/var/homebrew/linked/qt/5.9.9/clang_64/... and the build script is the only way I have ever fetched 5.9.9 on this Mac.

That's really strange. At least in this PR I moved away from homebrew. The path should also be /usr/local/opt/qt.

I don't fully understand it yet, but I also have a 5.12.9 under /usr/local/var/homebrew too, from the original build I did with this PR

Strange. Maybe you have two versions of Qt installed?

Have a look at this gh actions run: https://github.com/ann0see/jamulus/actions/runs/555358355

I just pushed the 5.9.9 version. Let's wait and see if it works.

Yes. It seems to work (although I haven't tested the binary yet.)

Strange. Maybe you have two versions of Qt installed?

The only way I have installed those two versions of Qt is by running the autorelease_mac.sh script, firstly the version in the PR with 5.12.9, and secondly after editing it to refer to 5.9.9, so I have to infer that either aqt is using homebrew behond the scenes, or the build process is.

That's really strange. I'm quite sure that no other script (at least from which I know) installs it via brew.
Did you use another version of Qt on your mac before running autorelease_mac.sh? Maybe this causes the problem?

OK - @ann0see do you want to change the CodeQL to use 5.9.9 in this same PR? That should be kept in line (I really wish it was just all part of the same work flow).

Hmm. I think @nefarius2001 is working on something (at least I saw some activity on his repo). Yes, CodeQl and this should both use the same version of Qt, I think. Can somebody please check if the produced binary on my repo works on older macOS versions?

true, I create #978 to give a preview

OK, I've tracked down where the homebrew stuff comes from on my Mac. I installed Qt 5.15.1 in October using brew, when I was testing out my work on audio packet ordering. That installed Qt into /usr/local/Cellar/qt/5.15.1 and it also set up the following symbolic links:

Tonys-MBP:jamulus tony$ ls -l /usr/local/var/homebrew/linked/qt /usr/local/opt/qt*
lrwxr-xr-x  1 tony  admin  19 13 Oct 15:04 /usr/local/opt/qt -> ../Cellar/qt/5.15.1
lrwxr-xr-x  1 tony  admin  19 13 Oct 15:04 /usr/local/opt/qt5 -> ../Cellar/qt/5.15.1
lrwxr-xr-x  1 tony  admin  25 13 Oct 15:04 /usr/local/var/homebrew/linked/qt -> ../../../Cellar/qt/5.15.1

Consequently, when aqt installed the older versions under /usr/local/opt/qt, it followed the existing symbolic link and installed them there. Since the homebrew link pointed to the same place, they appeared to be under there too, when doing a grep -r.

So the homebrew stuff is a red herring, but the installed versions of Qt under /usr/local/opt/qt do still have hard-coded references to /Users/qt/.... But as I said, that appears to be benign.

true, I create #978 to give a preview

Are we going to close this and proceed there?

Yes. I'll close this now.

@pljones if #978 takes too long for you (for the next release), we can also merge this one.

I'm not sure: since nefarius's PR is almost finished, we could just close this or do you want this to be merged?

Wait for now.