Some sort of convert bug? Started on June 18th

[IainNZ]

INFO: Installing Grid v0.3.2
INFO: Package database updated
ERROR: no method convert(Type{Array{Float64,1}}, Array{Float64,2})
 in ctranspose! at array.jl:1280
 in ctranspose at array.jl:1316
 in include at ./boot.jl:245
 in include_from_node1 at ./loading.jl:128
 in include at ./boot.jl:245
 in include_from_node1 at loading.jl:128
 in process_options at ./client.jl:285
 in _start at ./client.jl:354
while loading /home/idunning/pkgtest/.julia/v0.3/Grid/test/grid.jl, in expression starting on line 294
while loading /home/idunning/pkgtest/.julia/v0.3/Grid/test/runtests.jl, in expression starting on line 1
INFO: Package database updated

[tomasaschan]

Not caused by changes to Grid (we had no commits between May 23 and June 27) so I suspect this has to do with something in core Julia. I'll take a look and see if I can figure out what the problem is.

Edit: I think I found the problem, although I haven't located the change that broke it.

On line 294, which the error refers to, we define a = zeros(5,5). Or so we think - but a was already used on line 255 to hold a = zeros(5). Changing the variable name should fix this.

[tomasaschan]

By the way, is there a reason PkgEval didn't pick this up and file an issue automatically?

[IainNZ]

Re: why no auto-issue, same question occurred to me. I'm really not sure, probably an error on my part (its not fully automatic, I have to approve it, perhaps I forgot)

[tomasaschan]

@IainNZ Could it be that PkgEval only tests tagged versions? I don't think we've tagged anything since May either.

Well, I started running the test suite and switching out variable names, but there were quite a lot of other ones that didn't work out either. Lots of variables in the test script are named things like A, a, x etc, and re-used with no concerns about type stability. I think the best way out might be to encapsulate parts of the script in various functions and then call them all at the end.

Let's just make sure we don't tag a new version until this is fixed.

[IainNZ]

That test error is occuring on the tagged version - it simply uses Pkg.add(). People have been messing with this stuff in Base so I'm not too suprised something broke.

[timholy]

Oh, rats; I fixed this in db7cad0d75f7df4cdd2011c5eea3b10dff3fc8a1 but forgot to tag a new version. Fixed.

@IainNZ, this is part of what I was meaning when at JuliaCon I suggested that it would be great to re-run failing tests that same day to increase your confidence that the failures are real. I know if I were relying on my own brain to remember "failed today, let's see what happens tomorrow and if so I'll approve that email," I would forget 99% of the time. But I'm sure you're much better about this kind of thing than me :smile:.

[IainNZ]

Hah I was definitely thinking of that conversation when I noticed this!