about summary refs log tree commit homepage
path: root/KNOWN_ISSUES
diff options
context:
space:
mode:
Diffstat (limited to 'KNOWN_ISSUES')
-rw-r--r--KNOWN_ISSUES35
1 files changed, 28 insertions, 7 deletions
diff --git a/KNOWN_ISSUES b/KNOWN_ISSUES
index bc3dac5..2bd4151 100644
--- a/KNOWN_ISSUES
+++ b/KNOWN_ISSUES
@@ -3,16 +3,28 @@
 Occasionally odd {issues}[link:ISSUES.html] arise without a transparent or
 acceptable solution.  Those issues are documented here.
 
+* PRNGs (pseudo-random number generators) loaded before forking
+  (e.g. "preload_app true") may need to have their internal state
+  reset in the after_fork hook.  Starting with \Unicorn 3.6.1, we
+  have builtin workarounds for Kernel#rand and OpenSSL::Random users,
+  but applications may use other PRNGs.
+
+* Under some versions of Ruby 1.8, it is necessary to call +srand+ in an
+  after_fork hook to get correct random number generation.  We have a builtin
+  workaround for this starting with \Unicorn 3.6.1
+
+  See http://redmine.ruby-lang.org/issues/show/4338
+
+* On Ruby 1.8 prior to Ruby 1.8.7-p248, *BSD platforms have a broken
+  stdio that causes failure for file uploads larger than 112K.  Upgrade
+  your version of Ruby or continue using Unicorn 1.x/3.4.x.
+
 * For notes on sandboxing tools such as Bundler or Isolate,
   see the {Sandbox}[link:Sandbox.html] page.
 
-* Under Ruby 1.9.1, methods like Array#shuffle and Array#sample will
-  segfault if called after forking.  This is fixed in trunk (r26936) and
-  should be backported to the next 1.9.1 stable release (after p378).
-  Until then, it is advisable to call "Kernel.rand" in your after_fork
-  hook to reinitialize the random number generator.
-
-  See http://redmine.ruby-lang.org/issues/show/2962 for more details
+* nginx with "sendfile on" under FreeBSD 8 is broken when
+  uploads are buffered to disk.  Disabling sendfile is required to
+  work around this bug which should be fixed in newer versions of FreeBSD.
 
 * When using "preload_app true", with apps using background threads
   need to restart them in the after_fork hook because threads are never
@@ -22,6 +34,15 @@ acceptable solution.  Those issues are documented here.
   deadlocks.  The core Ruby Logger class needlessly uses a MonitorMutex
   which can be disabled with a {monkey patch}[link:examples/logger_mp_safe.rb]
 
+== Known Issues (Old)
+
+* Under Ruby 1.9.1, methods like Array#shuffle and Array#sample will
+  segfault if called after forking.  Upgrade to Ruby 1.9.2 or call
+  "Kernel.rand" in your after_fork hook to reinitialize the random
+  number generator.
+
+  See http://redmine.ruby-lang.org/issues/show/2962 for more details
+
 * Rails 2.3.2 bundles its own version of Rack.  This may cause subtle
   bugs when simultaneously loaded with the system-wide Rack Rubygem
   which Unicorn depends on.  Upgrading to Rails 2.3.4 (or later) is