Broken formulas cleanup

[sbesson]

This PR cleans up broken formulas in this tap and fix Travis. The zeroc-ice formula are dropped in favor of the official https://github.com/zeroc-ice/homebrew-tap tap. The omero formula are irreversibly broken given changes to the Homebrew build strategy.

The build should now be green and test the installation of the bioformats formula. Internally this also means we should be able to drop the Bio-Formats Homebrew job job and use exclusively Travis for validation.

/cc @dgault @joshmoore @jburel

[joshmoore]

Following on from https://www.openmicroscopy.org/community/viewtopic.php?f=5&t=7853#p19103 I tried brew install ome/alt/zeroc-ice35 --with-python this morning which fails with:

==> Downloading http://download.oracle.com/berkeley-db/db-5.3.28.tar.gz
Already downloaded: /Users/jamoore/Library/Caches/Homebrew/zeroc-ice35--berkeley-db-5.3.28.tar.gz
==> ../dist/configure --disable-debug --prefix=/usr/local/Cellar/zeroc-ice35/3.5.1_1/libexec --mandir=/usr/local/Cellar/zeroc-ice35/3.5
==> make install
==> make prefix=/usr/local/Cellar/zeroc-ice35/3.5.1_1 install_mandir=/usr/local/Cellar/zeroc-ice35/3.5.1_1/share/man/man1 install_slice
==> make install prefix=/usr/local/Cellar/zeroc-ice35/3.5.1_1 install_mandir=/usr/local/Cellar/zeroc-ice35/3.5.1_1/share/man/man1 insta
==> make prefix=/usr/local/Cellar/zeroc-ice35/3.5.1_1 install_mandir=/usr/local/Cellar/zeroc-ice35/3.5.1_1/share/man/man1 install_slice
==> make install prefix=/usr/local/Cellar/zeroc-ice35/3.5.1_1 install_mandir=/usr/local/Cellar/zeroc-ice35/3.5.1_1/share/man/man1 insta
==> make prefix=/usr/local/Cellar/zeroc-ice35/3.5.1_1 install_mandir=/usr/local/Cellar/zeroc-ice35/3.5.1_1/share/man/man1 install_slice
Last 15 lines from /Users/jamoore/Library/Logs/Homebrew/zeroc-ice35/07.make:
        if(PyTuple_SET_ITEM(args.get(), 0, (PyObject*)obj) < 0) // PyTuple_SET_ITEM steals a reference.
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^ ~
Operation.cpp:3665:54: error: ordered comparison between pointer and zero ('PyObject *' (aka '_object *') and 'int')
    if(PyTuple_SET_ITEM(args.get(), start, ip.get()) < 0)
       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^ ~
Operation.cpp:3677:56: error: ordered comparison between pointer and zero ('PyObject *' (aka '_object *') and 'int')
    if(PyTuple_SET_ITEM(args.get(), start, curr.get()) < 0)
       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^ ~
Operation.cpp:3697:60: error: ordered comparison between pointer and zero ('PyObject *' (aka '_object *') and 'int')
        if(PyTuple_SET_ITEM(args.get(), 0, (PyObject*)obj) < 0) // PyTuple_SET_ITEM steals a reference.
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ^ ~
15 errors generated.
make[2]: *** [Operation.o] Error 1
make[1]: *** [all] Error 1
make: *** [all] Error 1

If reporting this issue please do so at (not Homebrew/brew or Homebrew/core):
https://github.com/ome/homebrew-alt/issues

Happy to file this as an issue but equally happy to remove the broken and dependent formulae here if we're dropping testing from travis.

[sbesson]

My understanding is that many of our formulas are broken for different reasons. Proposal would be to move forward with the cleanup suggested in the description and @joshmoore's comment with the following rationales:

• zeroc-ice{33,34,35}: OME maintaining the distribution for an external dependency was really a workaround in the early days. Propose to remove in favor of https://github.com/zeroc-ice/homebrew-tap since the ZeroC team is now maintaining its own tap.
• berkeley-db{44,45} are dependencies of zeroc-ice and the ZeroC tap also implements them.
• omero{43,44,50,51,52,53} are broken since Homebrew switched to impose sandboxing for building (so the ~/.m2 directory cannot be used)

[joshmoore]

Quite the cleanup, but 💚 Assuming the disappearance won't cause any unintended side-effects, I think this is (unfortunately) our best choice currently.

[dgault]

The Travis build for Bio-Formats is now working as expected and the Bio-Formats install completes successfully with the installation of the correct version (5.8.0). The Bio-Formats test completes as expected with showinf and bfconvert tested. The updated documentation also looks good.

As far as removing the broken OMERO, zeroc-ice and berkley-db formulas, this is fine from my perspective if there is no easy way of getting the OMERO formulas working currently.

PR is good from my side.

[sbesson]

Thanks all. Merging and actioning the removal of the CI jobs.