-
Evan Phoenix committed
Nov 19, 2010 -
Fixed Bignum#<=> with an object instance.
Brian Ford authored and Evan Phoenix committedNov 19, 2010 -
Fix Debugger.start in ruby-debug compat mode
Evan Phoenix committedNov 19, 2010
-
Permit characters after (rbx|rubinius) in --prefix dir.
Brian Ford committedNov 17, 2010 This solves an issue where a --prefix dir like rbx-1.1.1-junk already exists and the user gives that dir to configure again. In that case, we assume it to be a Rubinius install dir and install into it.
-
Match an empty string properly. Fixes #573
Evan Phoenix committedNov 17, 2010 -
Add spec for matching an empty string
Evan Phoenix committedNov 17, 2010 -
Add Method#source, show source in Method#inspect
Evan Phoenix committedNov 17, 2010
-
Evan Phoenix committed
Nov 16, 2010
-
Make CM#locate_line more strict. Fixes #563
Evan Phoenix committedNov 15, 2010 -
Make sure that the preinstalled dir is always present
Evan Phoenix committedNov 15, 2010 -
Updated Rapa source to 36f91258.
Brian Ford committedNov 15, 2010 -
Fixed code confusing GCC overflow checks in unpack.
Brian Ford committedNov 15, 2010 -
Updated Rapa source to e00f5a3a.
Brian Ford committedNov 15, 2010 -
Replaced size_t for native_int in pack, unpack.
Brian Ford committedNov 15, 2010 -
Evan Phoenix committed
Nov 15, 2010 -
Merge remote branch 'jeremydurham/master'
Evan Phoenix committedNov 15, 2010 -
Show the missing file error to -c more clearly. Fixes #568
Evan Phoenix committedNov 15, 2010 -
Evan Phoenix committed
Nov 15, 2010 -
Added spec to test Float#/ properly coercing other object
jeremydurham committedNov 15, 2010 -
Use redo_coerced when calling Float#divide. Fixes #564
jeremydurham committedNov 15, 2010 -
Added irc and github information
jeremydurham committedNov 15, 2010
-
We don't use anything deprecated, so -Wno-deprecated isn't needed
dbussink committedNov 14, 2010 Verified that we can build without this option on OS X 10.6, Debian Lenny 64 and FreeBSD 8.1. Seems safe to remove this option, if it turns out we need it for some obscure platform it can be added back.
-
Remove old SendSite/Selector cruft
Evan Phoenix committedNov 8, 2010 -
Fix CompiledMethod#locate_line
Evan Phoenix committedNov 8, 2010 -
Add Rubinius::OS_ARGV and Rubinius::OS_STARTUP_DIR
Evan Phoenix committedNov 8, 2010 -
Change IO::Buffer => IO::InternalBuffer. Fixes #560
Evan Phoenix committedNov 8, 2010 -
Make sure there is a fallback policy. Fixes #561
Evan Phoenix committedNov 8, 2010 -
Explicitly add -fno-omit-frame-pointer to the build flags
dbussink committedNov 8, 2010 On FreeBSD -O2 adds -fomit-frame-pointer which breaks execinfo.h usage. We explicitly disable omitting frame pointers, which is also the behavior on other platforms such as OS X. Here -O2 does not imply omitting frame pointers, so adding this also doesn't negatively impact performance there.
-
Workaround for FreeBSD by exposing dlopen directly
dbussink committedNov 8, 2010 Apparently the dlopen handle that is retrieved through dlsym() refers to a non working implementation. We expose the pointer therefore directly so it also works on FreeBSD. Also see the following link where the faulty dlopen code is shown: http://unix.derkeiler.com/Mailing-Lists/FreeBSD/current/2008-10/msg00237.html
-
Support for native extensions with a . in their name
Tony Arcieri committedNov 7, 2010 On MRI, if you have a library named "foo.bar", it strips the .bar component out of the name when determining which Init function to call. So in the case of a library named "foo.bar", MRI calls "Init_foo" This change matches the MRI behavior for determining which Init function to call
-
Explicitly mark one pipe as write only and one as read only
dbussink committedNov 7, 2010 When connecting two pipes, we create them with explicit modes. This fixes a spec failure on FreeBSD where the pipes were created with a different mode. This is the same behavior that MRI has.
-
Add hook to allow code to install a custom debugger
Evan Phoenix committedNov 6, 2010
-
Only shutdown spec HTTP server at exit.
Brian Ford committedNov 5, 2010 Due to system timing sensitivities, we were getting races on both startup and shutdown. There does not appear to be any reason not to leave the server running until the process exits. If there are problems that arise, we'll revisit this.
-
Add ability to scope and stop only in certain stack frames. Fixes #558
Evan Phoenix committedNov 5, 2010 -
dbussink committed
Nov 5, 2010