Re: GNU Guile 2.1.3 released

Re: GNU Guile 2.1.3 released

[Daniel Llorens]

>> On Jun 19, 2016, at 2:31 AM, Andy Wingo <wingo@pobox.com> wrote:
>> 
>> We are pleased to announce GNU Guile release 2.1.3.
> 
> Tried on my Mac OS 10.10.5 against Macports 2.3.4 w/ gcc-mp-5.  
> configure went OK
> make went OK
> make check generated FAIL for test-language and test-stack-overflow and stopped working after test-stack-overflow.
> 
> See attached.
> 
> Matt

I can confirm the failure of test-stack-overflow on 10.9.5 too. 

Re: GNU Guile 2.1.3 released

[Andy Wingo]

On Thu 23 Jun 2016 14:59, Daniel Llorens <daniel.llorens@bluewin.ch> writes:

>>> We are pleased to announce GNU Guile release 2.1.3.
>> 
>> Tried on my Mac OS 10.10.5 against Macports 2.3.4 w/ gcc-mp-5.  
>> make check generated FAIL for test-language and test-stack-overflow and stopped working after test-stack-overflow.
>
> I can confirm the failure of test-stack-overflow on 10.9.5 too. 

I am not sure how to debug this one as I don't have a Mac.  If you are
able to figure this one out, Daniel, that would be a great help.

Andy

Re: GNU Guile 2.1.3 released

[Matt Wette]

[-- Attachment #1: Type: text/plain, Size: 1539 bytes --]


> On Jun 23, 2016, at 9:11 AM, Andy Wingo <wingo@pobox.com> wrote:
> 
> On Thu 23 Jun 2016 14:59, Daniel Llorens <daniel.llorens@bluewin.ch> writes:
> 
>>>> We are pleased to announce GNU Guile release 2.1.3.
>>> 
>>> Tried on my Mac OS 10.10.5 against Macports 2.3.4 w/ gcc-mp-5.  
>>> make check generated FAIL for test-language and test-stack-overflow and stopped working after test-stack-overflow.
>> 
>> I can confirm the failure of test-stack-overflow on 10.9.5 too. 
> 
> I am not sure how to debug this one as I don't have a Mac.  If you are
> able to figure this one out, Daniel, that would be a great help.

I was able to get gdb-uninstalled-guile to run (by symlinking libguile-2.2.dynlib in $prefix/lib).  I accidentally killed the process.  But before I restart I want to understand the intended behavior.  The relevant code from test-stack-overflow is this:

;; 100 MB.
(define *limit* (* 100 1024 1024))

(call-with-values (lambda () (getrlimit 'as))
  (lambda (soft hard)
    (unless (and soft (< soft *limit*))
      (setrlimit 'as (if hard (min *limit* hard) *limit*) hard))))

(define (test)
  (catch 'stack-overflow
    (lambda ()
      (let lp ()
        (lp)
        (error "should not be reached")))
    (lambda _
      #t)))

On my mac (getrlimit ‘as) => #f #f.   So it looks like the args to setrlimit are 100 MB and #f.  Is the program supposed to die gracefully at 100 MB?   When I (accidentally) stopped the program (under gdb) the process was using 56 GB of memory.

Matt

[-- Attachment #2: Type: text/html, Size: 3822 bytes --]

Re: GNU Guile 2.1.3 released

[Daniel Llorens]

[-- Attachment #1: Type: text/plain, Size: 2353 bytes --]


On 24 Jun 2016, at 18:18, Matt Wette <matt.wette@gmail.com> wrote:

> 
>> On Jun 23, 2016, at 9:11 AM, Andy Wingo <wingo@pobox.com> wrote:
>> 
>> On Thu 23 Jun 2016 14:59, Daniel Llorens <daniel.llorens@bluewin.ch> writes:
>> 
>>>>> We are pleased to announce GNU Guile release 2.1.3.
>>>> 
>>>> Tried on my Mac OS 10.10.5 against Macports 2.3.4 w/ gcc-mp-5.  
>>>> make check generated FAIL for test-language and test-stack-overflow and stopped working after test-stack-overflow.
>>> 
>>> I can confirm the failure of test-stack-overflow on 10.9.5 too. 
>> 
>> I am not sure how to debug this one as I don't have a Mac.  If you are
>> able to figure this one out, Daniel, that would be a great help.
> 
> I was able to get gdb-uninstalled-guile to run (by symlinking libguile-2.2.dynlib in $prefix/lib).  I accidentally killed the process.  But before I restart I want to understand the intended behavior.  The relevant code from test-stack-overflow is this:
> 
> ;; 100 MB.
> (define *limit* (* 100 1024 1024))
> 
> (call-with-values (lambda () (getrlimit 'as))
>   (lambda (soft hard)
>     (unless (and soft (< soft *limit*))
>       (setrlimit 'as (if hard (min *limit* hard) *limit*) hard))))
> 
> (define (test)
>   (catch 'stack-overflow
>     (lambda ()
>       (let lp ()
>         (lp)
>         (error "should not be reached")))
>     (lambda _
>       #t)))
> 
> On my mac (getrlimit ‘as) => #f #f.   So it looks like the args to setrlimit are 100 MB and #f.  Is the program supposed to die gracefully at 100 MB?   When I (accidentally) stopped the program (under gdb) the process was using 56 GB of memory.
> 
> Matt

Yes, the OS is supposed to send a signal when the limit is reached and Guile handles it and throws an exception. So the test just catches that a few times and then ends.

We looked at this over IRC and it seems OS X simply ignores these limits. There are complaints about it all over the web.  If you hadn't stopped the process it would have started to eat swap space and eventually take down your machine. 

For the time being these tests are disabled in master (http://git.savannah.gnu.org/gitweb/?p=guile.git;a=commitdiff;h=9687334ff52a2def369e7abb8563401351db9be7).

Thanks for bringing this up.  I use Guile on OS X but I hadn't run make check for a while.






[-- Attachment #2: Type: text/html, Size: 5061 bytes --]

Re: GNU Guile 2.1.3 released

[Andy Wingo]

On Fri 24 Jun 2016 18:18, Matt Wette <matt.wette@gmail.com> writes:

> On my mac (getrlimit ‘as) => #f #f. So it looks like the args to
> setrlimit are 100 MB and #f. Is the program supposed to die gracefully
> at 100 MB? When I (accidentally) stopped the program (under gdb) the
> process was using 56 GB of memory.

Apparently setrlimit(RLIMIT_AS, ...) "succeeds" on macOS but doesn't
actually do anything :-//////

Daniel added a check for macOS in the test case and so we won't test
setrlimit.

Andy