JFYI seg fault while building

['John Clements' via dev-redirect]

Just got a seg fault while running setup, apparently during a call to fast_equal_hash_key.

Here’s a gist of the crash report:

https://gist.github.com/jbclements/a8601a360881b5826138

Here’s a gist of the terminal output:

https://gist.github.com/jbclements/b2c262c81ef36ab01377

I’m building commit dfab18fe4729e71f53d0056572e3f85a72467247

(though in principle this could depend on the commit that all of my cloned trees are on, too…)

I’m just going to try building again.

John

['John Clements' via dev-redirect]

I saw this crash again (thread 8 crashed during raco setup, again during a call to fast_equal_hash_key).

Crash report on request.

John

> --
> You received this message because you are subscribed to the Google Groups "Racket Developers" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to racket-dev+...@googlegroups.com.
> To post to this group, send email to racke...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/racket-dev/7ccdc730-097a-4d8d-ae6e-3804bc7f9de1%40mtasv.net.
> For more options, visit https://groups.google.com/d/optout.

[Matthew Flatt]

I've finally been able to track down this bug. It's due to a misapplied
unsafe operation in the Typed Racket optimizer, and that bug is
triggered when compiling "midi-readwrite" --- which is why it's so
consistent for you and not seen by others who don't have the package
installed.

In the TR implementation, changing `unsafe-Req-seq` to use `Rep-seq`
instead of `unsafe-struct*-ref` produces the error

Rep-seq: contract violation
expected: Rep?
given: #f
compilation context...:
..../pkgs/midi-readwrite/midi-readwrite/midi-read.rkt
context...:
..../typed-racket/types/subtype.rkt:238:0: subtype*
....

[Vincent St-Amour]

Sounds like a bug in the TR optimizer. I'll look into it.

Vincent

> To view this discussion on the web visit https://groups.google.com/d/msgid/racket-dev/56357047.a1fb440a.8cd5e.ffffc2d6SMTPIN_ADDED_MISSING%40mx.google.com.

[Sam Tobin-Hochstadt]

If it's an eror with `Rep-seq` being passed #f, then probably it's
just a bug somewhere else in TR, rather than related to the optimizer,
unless it goes away when the optimizer is off.

Sam

On Sun, Nov 1, 2015 at 8:57 AM, Vincent St-Amour
<stam...@eecs.northwestern.edu> wrote:
> Sounds like a bug in the TR optimizer. I'll look into it.
>
> Vincent
>
>
> On Sat, 31 Oct 2015 20:52:01 -0500,
> Matthew Flatt wrote:
>>

>> To view this discussion on the web visit https://groups.google.com/d/msgid/racket-dev/56357047.a1fb440a.8cd5e.ffffc2d6SMTPIN_ADDED_MISSING%40mx.google.com.

>> For more options, visit https://groups.google.com/d/optout.
>

> --
> You received this message because you are subscribed to the Google Groups "Racket Developers" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to racket-dev+...@googlegroups.com.
> To post to this group, send email to racke...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/racket-dev/m2egg9ddq7.wl-stamourv%40eecs.northwestern.edu.