mkdir: supply path to SystemError

[timholy]

Some mkdir errors are very hard to debug unless you know what path it was trying to create. For example:

INFO: Upgrading Escher: v0.1.0 => v0.2.1
INFO: Upgrading Mux: v0.1.1 => v0.2.0
INFO: Installing Plots v0.4.2
INFO: Removing Color v0.2.11
INFO: Rolling back install of Plots
INFO: Rolling back Mux from v0.2.0 to v0.1.1
INFO: Rolling back Escher from v0.2.1 to v0.1.0
ERROR: SystemError: mkdir: File exists
 in mkdir at file.jl:42
 in cptree at file.jl:100
 in cp at file.jl:120
 in rename at fs.jl:121
 in remove at pkg/write.jl:50
 in resolve at ./pkg/entry.jl:459
 in resolve at pkg/entry.jl:404
 in anonymous at pkg/entry.jl:198
 in transact at pkg/git.jl:97
 in _checkout at pkg/entry.jl:190
 in free at pkg/entry.jl:220
 in anonymous at pkg/dir.jl:31
 in cd at file.jl:22
 in cd at pkg/dir.jl:31
 in free at pkg.jl:40

The problem turned out to be a symlink ~/.julia/v0.4/.trash/Color (I was using symlinks to facilitate testing changes on both 0.3 and 0.4), but until I modified the error to show the path this was difficult to figure out.

The new version prints something like this:

"SystemError (with /home/tim/.julia/v0.4/.trash/Color): mkdir: File exists"

[tkelman]

The keyword argument should be documented here.

[tkelman]

Oh hah, I was having a hard time debugging something for exactly this reason (on release-0.4 though). Why did you go with a keyword argument instead of constructing the message as the first input to systemerror ? We could use similar treatments for most other operations that can throw SystemErrors.