Example of a Package Install Breaking

75 views
Skip to first unread message

Leah Hanson

unread,
Dec 27, 2014, 10:32:50 PM12/27/14
to Julia Users
I'm working on a book chapter about using the Julia package manager. One example I want to include is debugging a broken package install. I'm currently expecting to focus on a problem caused by binary installs, since that seems to be the mostly likely cause of a Pkg.add failing.

However, it's hard for me to conjure one at will. Does anyone here have any suggestions?

It would be helpful if you either (A) have a story about a package install that broke at first and that you then got working (please say what happened and how you fixed it) or (B) an idea for how to reliably create a problem for Pkg.add (which would make it easier for me to get the example output for the book).

Thanks!
Leah

Tim Holy

unread,
Dec 27, 2014, 11:42:46 PM12/27/14
to julia...@googlegroups.com
Images has a somewhat overwhelming number of closed issues (and likely, a few
still open) regarding problems with ImageMagick. Take your pick :-).

--Tim

Stefan Karpinski

unread,
Dec 28, 2014, 12:01:22 AM12/28/14
to julia...@googlegroups.com
I'm somewhat thrilled that the problem here is the inability to conjure a failing case :)

Avik Sengupta

unread,
Dec 28, 2014, 9:57:09 AM12/28/14
to julia...@googlegroups.com


On Sunday, 28 December 2014 05:01:22 UTC, Stefan Karpinski wrote:
I'm somewhat thrilled that the problem here is the inability to conjure a failing case :)

I'm not sure. I've had the feeling that OSX has had the worst experience in installing binary dependencies.

One example I had is last week: https://github.com/JuliaWeb/GnuTLS.jl/issues/25 . The issue is still open. It may or may not have been fixed with a new Bindeps release.

Current status on my OSX Mavericks, Julia 0.3.4 on Pkg.build(Cairo)

==> Summary
🍺  /Users/aviks/.julia/v0.3/Homebrew/deps/usr/Cellar/icu4c/52.1: 239 files, 60M
==> Installing pango dependency: staticfloat/juliadeps/harfbuzz
==> Downloading https://juliabottles.s3.amazonaws.com/harfbuzz-0.9.33.mavericks.bottle.tar.gz
Already downloaded: /Users/aviks/Library/Caches/Homebrew.jl/harfbuzz-0.9.33.mavericks.bottle.tar.gz
Error: SHA1 mismatch
Expected: 3d136380b810141a5d35fda62bdd35b8771f9314
Actual: a97e9fdfaa2dd88f567f82933f9b27384d822f47
Archive: /Users/aviks/Library/Caches/Homebrew.jl/harfbuzz-0.9.33.mavericks.bottle.tar.gz
To retry an incomplete download, remove the file above.
Warning: Bottle installation failed: building from source.
Error: /Users/aviks/.julia/v0.3/Homebrew/deps/usr/opt/pkg-config not present or broken
Please reinstall pkg-config. Sorry :(
======================================================================[ ERROR: Cairo ]======================================================================

failed process: Process(`/Users/aviks/.julia/v0.3/Homebrew/deps/usr/bin/brew install --force-bottle staticfloat/juliadeps/pango`, ProcessExited(1)) [1]
while loading /Users/aviks/.julia/v0.3/Cairo/deps/build.jl, in expression starting on line 144

============================================================================================================================================================

======================================================================[ BUILD ERRORS ]======================================================================

WARNING: Cairo had build errors.

 - packages with build errors remain installed in /Users/aviks/.julia/v0.3
 - build the package(s) and all dependencies with `Pkg.build("Cairo")`
 - build a single package by running its `deps/build.jl` script

============================================================================================================================================================


Note that a couple of weeks ago I was getting a different error in installing Cairo.

I must admit that I have NEVER been able to install TK  or Cairo packages in Julia without manual intervention. Typically one tends to find a manual workaround, which lasts till the next OS or machine upgrade.

Regards
-
Avik

 

Tim Holy

unread,
Dec 28, 2014, 10:23:52 AM12/28/14
to julia...@googlegroups.com
On Sunday, December 28, 2014 12:01:10 AM Stefan Karpinski wrote:
> I'm somewhat thrilled that the problem here is the inability to conjure a
> failing case :)

Deservedly so!

--Tim
Reply all
Reply to author
Forward
0 new messages