From @labster

01​:56 labster m​: "/".IO.mkdir # well that's a surprise
01​:56 camelia rakudo-moar b38385​: OUTPUT«(signal SEGV)»
01​:58 labster j​: "/".IO.mkdir
01​:58 camelia rakudo-jvm 6c0f93​: OUTPUT«Failed to create directory
'/' with mode '0o777'​: File / already exists␤ in block <unit> at
/tmp/LFlB2yNApP​:1␤␤Actually thrown at​:␤ in block <unit> at
/tmp/LFlB2yNApP​:1␤␤»
01​:58 labster looks like moarbugs
Also​:
brent@​ragnar ~/c/roast (master)> perl6 -e 'mkdir("/")'
fish​: Job 1, 'perl6 -e 'mkdir("/")'' terminated by signal SIGSEGV (Address
boundary error)

Also segfaults on OS X. JVM behavior looks OK.

The upside is that Perl 6 is so terse, you can segfault in only 10
characters!

From @smls

No longer segfaults with current r-m​:

  ➜ "/".IO.mkdir;
  Failed to create directory '/' with mode '0o777'​: Failed to mkdir​: 2

Ticket should be closable once a test has been added to Roast.

The RT System itself - Status changed from 'new' to 'open'

From @zoffixznet

Ticket should be closable once a test has been added to Roast.

Is it really appropriate to attempt to create such directories? I'm hesitant to have roast touch things outside of its directory.

From @zoffixznet

Tests added​: Raku/roast@423ee643ea

@zoffixznet - Status changed from 'open' to 'resolved'

From @niner

Without having checked this, it looks to me like the bug was not about trying to create the root directory, but a failure to handle a permission problem trying to create a directory.

If that's true, then the test should easily be able to re-create the circumstances. Just create a temporary directory and remove your own write permissions. Then try to create a subdirectory which should fail of course. Then you can remove the temporary directory again.

From @geekosaur

On Mon, Jul 4, 2016 at 6​:44 PM, Zoffix Znet via RT <
perl6-bugs-followup@​perl.org> wrote​:

Ticket should be closable once a test has been added to Roast.
Is it really appropriate to attempt to create such directories? I'm
hesitant to have roast touch things outside of its directory.

It's a directory that must always exist and therefore always fail in a
consistent way.

--
brandon s allbery kf8nh sine nomine associates
allbery.b@​gmail.com ballbery@​sinenomine.net
unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net

From zoffix@zoffix.com

And when a bug creeps in and destroy's someone's system? We already
destroyed one user's entire system with a bug in panda. It wasn't a
pretty picture.

It's fine to check failure modes on an existing directory we create
during testing, but I'm a definite -1 on attempting to much around
with '/'

Cheers,
ZZ

Quoting Brandon Allbery <allbery.b@​gmail.com>​:

On Mon, Jul 4, 2016 at 6​:44 PM, Zoffix Znet via RT <
perl6-bugs-followup@​perl.org> wrote​:

Ticket should be closable once a test has been added to Roast.
Is it really appropriate to attempt to create such directories? I'm
hesitant to have roast touch things outside of its directory.

It's a directory that must always exist and therefore always fail in a
consistent way.

--
brandon s allbery kf8nh sine nomine associates
allbery.b@​gmail.com ballbery@​sinenomine.net
unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net

From zoffix@zoffix.com

Or we could have some extra tests category for dangerous tests.

I know make stresstest does extra tests. Would this test fit into those?

Quoting zoffix@​zoffix.com​:

And when a bug creeps in and destroy's someone's system? We already
destroyed one user's entire system with a bug in panda. It wasn't a
pretty picture.

It's fine to check failure modes on an existing directory we create
during testing, but I'm a definite -1 on attempting to much around
with '/'

Cheers,
ZZ

Quoting Brandon Allbery <allbery.b@​gmail.com>​:

On Mon, Jul 4, 2016 at 6​:44 PM, Zoffix Znet via RT <
perl6-bugs-followup@​perl.org> wrote​:

Ticket should be closable once a test has been added to Roast.
Is it really appropriate to attempt to create such directories? I'm
hesitant to have roast touch things outside of its directory.

It's a directory that must always exist and therefore always fail in a
consistent way.

--
brandon s allbery kf8nh sine nomine associates
allbery.b@​gmail.com ballbery@​sinenomine.net
unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net

From @geekosaur

On Tue, Jul 5, 2016 at 12​:06 AM, <zoffix@​zoffix.com> wrote​:

It's fine to check failure modes on an existing directory we create during
testing, but I'm a definite -1 on attempting to much around with '/'

The only way that would misbehave would indicate a system already corrupted
beyond recovery before the test was run.

--
brandon s allbery kf8nh sine nomine associates
allbery.b@​gmail.com ballbery@​sinenomine.net
unix, openafs, kerberos, infrastructure, xmonad http://sinenomine.net