[racket] Permission errors with nightlies?

Yes, it's a problem with yesterday's snapshot from Utah, and the most
recent one failed for unrelated reasons. I've kick off a new snapshot,
so it should be available in about 3 hours.

If you can tolerate errors provoked by bad snapshots, then I think
building against the snapshot is a good choice. I doubt that the error
rate has been as low as 1%, though (as different problems spanning two
days illustrate). Also, we have two snapshot sites to compensate for
failures at either one, and that works for manual downloads, but I
don't know how you'd take advantage of the two sites for your setting.
Of course, as you suggest, you should use the release if you need more
reliability.

At Tue, 4 Nov 2014 11:05:26 -0500, Greg Hendershott wrote:
> Background: https://travis-ci.org/greghendershott/frog/jobs/39956106
> 
> I'm using 
> http://www.cs.utah.edu/plt/snapshots/current/installers/racket-current-x86_64-li
> nux-precise.sh
> on Travis CI.
> 
> Suddenly I'm seeing dozens of error like this:
> 
> open-output-file: cannot open output file
>   path: 
> /usr/racket/share/pkgs/syntax-color-lib/syntax-color/compiled/lexer-contract_rkt
> .zo
>   system error: Permission denied; errno=13
>   context...:
>    /usr/racket/collects/compiler/cm.rkt:185:4
>    /usr/racket/collects/compiler/cm.rkt:470:0: maybe-compile-zo
>    /usr/racket/collects/compiler/cm.rkt:581:2: do-check
>    /usr/racket/collects/compiler/cm.rkt:620:15
>    /usr/racket/collects/compiler/../racket/private/map.rkt:113:23: loop
>    /usr/racket/collects/compiler/cm.rkt:581:2: do-check
>    /usr/racket/collects/compiler/cm.rkt:620:15
>    /usr/racket/collects/compiler/../racket/private/map.rkt:113:23: loop
>    /usr/racket/collects/compiler/cm.rkt:581:2: do-check
>    /usr/racket/collects/compiler/cm.rkt:226:11: for-loop
>    /usr/racket/collects/compiler/cm.rkt:470:0: maybe-compile-zo
>    /usr/racket/collects/compiler/cm.rkt:581:2: do-check
>    /usr/racket/collects/compiler/cm.rkt:696:4: compilation-manager-load-handler
>    standard-module-name-resolver
>    /usr/racket/collects/compiler/cm.rkt:309:0: compile-zo*
>    /usr/racket/collects/compiler/cm.rkt:512:26...
> 
> ISTR this happening a month or two ago, and it resolved itself on the
> next nightly build.
> 
> Now that it's happened again, I wanted to ask why it happens?
> 
> I like to have Travis CI build my projects against HEAD (nightlies) as
> well as released versions. I understand that nightlies might be buggy.
> So one answer might be, I should stop doing this. On the other hand I
> like the idea of an "early alert" if some change to Racket isn't
> compatible with something I'm doing. And 99% of the time it works
> fine. So I guess I'm asking about this specific 1% case, now that it's
> recurred.
> ____________________
>   Racket Users list:
>   http://lists.racket-lang.org/users