[#64517] Fw: Re: Ruby and Rails to become Apache Incubator Project — Tetsuya Kitahata <[email protected]>

What do you think? >> Ruby developers

13 messages 2014/08/23

[#64615] [ruby-trunk - Feature #10181] [Open] New method File.openat() — oss-ruby-lang@...

Issue #10181 has been reported by Technorama Ltd..

10 messages 2014/08/28
[#64616] Re: [ruby-trunk - Feature #10181] [Open] New method File.openat() — Eric Wong <normalperson@...> 2014/08/28

I like this feature.

[#64671] Fwd: [ruby-changes:35240] normal:r47322 (trunk): symbol.c (rb_sym2id): do not return garbage object — SASADA Koichi <ko1@...>

Why this fix solve your problem?

9 messages 2014/08/30
[#64672] Re: Fwd: [ruby-changes:35240] normal:r47322 (trunk): symbol.c (rb_sym2id): do not return garbage object — SASADA Koichi <ko1@...> 2014/08/30

(2014/08/30 8:50), SASADA Koichi wrote:

[ruby-core:64300] [ruby-trunk - Bug #9914] posix_fadvise() does not work correctly with _LARGE_FILES on 32-bit AIX

From: nagachika00@...
Date: 2014-08-11 13:44:04 UTC
List: ruby-core #64300
Issue #9914 has been updated by Tomoyuki Chikanaga.

Backport changed from 2.0.0: REQUIRED, 2.1: REQUIRED to 2.0.0: REQUIRED, 2.1: DONE

Backported into `ruby_2_1` branch at r47139.

----------------------------------------
Bug #9914: posix_fadvise() does not work correctly with _LARGE_FILES on 32-bit AIX
https://bugs.ruby-lang.org/issues/9914#change-48282

* Author: Rei Odaira
* Status: Closed
* Priority: Normal
* Assignee: Yutaka Kanemoto
* Category: platform/aix
* Target version: 
* ruby -v: ruby 2.2.0dev (2014-06-02 trunk 45270) [powerpc-aix7.1.0.0]
* Backport: 2.0.0: REQUIRED, 2.1: DONE
----------------------------------------
`test_advise` in `test/ruby/test_io.rb` fails on AIX.

~~~
[  1/170] TestIO#test_advise = 0.01 s
  1) Error:
TestIO#test_advise:
Errno::EINVAL: Invalid argument - /tmp/test_io20140606-7733390-1xgpc07
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2437:in `advise'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2437:in `block (4 levels) in test_advise'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2436:in `open'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2436:in `block (3 levels) in test_advise'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2435:in `each'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2435:in `block (2 levels) in test_advise'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2434:in `each'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2434:in `block in test_advise'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:1442:in `make_tempfile'
    /ss/home/rayod/Dev/Contribution/ruby-trunk-blue1/test/ruby/test_io.rb:2432:in `test_advise'
~~~
It is very strange, but the current AIX does not seem to support a 32-bit call to `posix_fadvise()` when `_LARGE_FILES` is defined.
http://www-01.ibm.com/support/docview.wss?uid=isg1IV56170

Until it is supported, we need this ugly patch to not call `posix_fadvise()`....

~~~diff
--- io.c        (revision 46310)
+++ io.c        (working copy)
@@ -8586,7 +8586,10 @@
     off = NIL_P(offset) ? 0 : NUM2OFFT(offset);
     l   = NIL_P(len)    ? 0 : NUM2OFFT(len);
 
-#ifdef HAVE_POSIX_FADVISE
+    /* AIX currently does not support a 32-bit call to posix_fadvise()
+     * if _LARGE_FILES is defined.
+     */
+#if defined(HAVE_POSIX_FADVISE) && !(defined(_AIX) && defined(_LARGE_FILES) && !defined(_ARCH_PPC64))
     return do_io_advise(fptr, advice, off, l);
 #else
     ((void)off, (void)l);      /* Ignore all hint */

~~~

FYI, detailed explanation follows:
The 2nd and 3rd arguemnts to `posix_fadvise()` are the `off_t` type. If `_LARGE_FILES` is defined, `off_t` is defined as `long long int` on the caller side. This means the 2nd and 3rd arguments are passed using two registers each on a 32-bit environment. However, the callee side does not support `_LARGE_FILES`, so `off_t` is assumed to be `long int`, which is passed using one register. As a result, the callee side can not receive a correct value for the 4th argument, `int advice`, throwing EINVAL.



-- 
https://bugs.ruby-lang.org/

In This Thread

Prev Next