Commits

Branch selector

 master

User selector

All users

All time

Commit History

Commits on Sep 30, 2013

• Finalized what-remains-to-do/

  lucasaiu

  committed
  323378a

• Fixed the very last bug: extensible buffers now have a "shape" to share among contexts

  lucasaiu

  committed
  8e6c3e3

Commits on Sep 18, 2013

• Lots of fixes: now opt.opt can be built on i386 and other non-multiruntime architectures

  lucasaiu

  committed
  21eab0f

Commits on Sep 16, 2013

• Removed some generated configuration files which shouldn't be under git control

  lucasaiu

  committed
  24b551e

• Fixed a nasty bug in selectgen.ml that I introduced last weel ( Printf.fprintf stderr "QQQQQQQQ %s\n%!" s;)

  lucasaiu

  committed
  39c2583

• Conditionally moved some contextual C globals back out of the struct, if they are accessed by the assembly runtime in some ports; this lets us keep existing assembly runtimes with no changes

  luca

  committed
  68c37b3

• Minor fixes in the i386 GNU/Linux backend, for compatibility with the multicontext system

  luca

  committed
  0a0c1b3

Commits on Sep 12, 2013

• Stuff

  lucasaiu

  committed
  03ef07b

• Bugfixes and stuff

  Luca Saiu

  committed
  c226911

Commits on Sep 10, 2013

• Fixed native code support for non-multiruntime architectures (developed on PowerPC). Fixed PowerPC assembly runtime with a cute little hack, which unfortunately will have to go.

  Luca Saiu

  committed
  170eefa

Commits on Sep 9, 2013

• Configuration improvements: now multi-context can be supported but still disabled in an installation, and configuration macros reflect this duality

  lucasaiu

  committed
  0250279

Commits on Sep 4, 2013

• Merge: trivial stuff

  lucasaiu

  committed
  3b7a1a7

• Trivial changes

  lucasaiu

  committed
  96de9f5

• Fixes for 32-bit and big-endian architectures, after testing on my old PowerPC machine

  Luca Saiu

  committed
  7abceb8

Commits on Sep 3, 2013

• Now the multiruntime system can be disabled at configuration time

  lucasaiu

  committed
  355cf08

Commits on Aug 6, 2013

• Two important changes: 1) changed the destruction of the ticker threads at systhread context finalization to be synchronous: this is important for ensuring that the masterlock is not used after con…

  lucasaiu

  committed
  486b53b

• In the previous commit message I forgot to mention that I also implemented Context.at_exit, and updated the Callback machinery so that its state is correctly cloned at split time

  lucasaiu

  committed
  24ef428

• Implemented context finalization with reference counting. It works with no threads and with systhread, but not yet (if ever) with vmthreads

  lucasaiu

  committed
  3aeef97

Commits on Jul 25, 2013

• Added a thin abstraction layer over sem_wait and sem_post, because sem_wait was being interrupted by signals in the bytecode runtime; now I correctly call it in a loop

  lucasaiu

  committed
  11901cb

Commits on Jul 23, 2013

• Fixed channel locking in the case of multi-context concurrent channel access without threads, and vmthreads. The vmthreads part is ugly and inefficient in case of contention over busy channels

  lucasaiu

  committed
  4de896d

Commits on Jul 16, 2013

• Adapted VM threads to multi-context; threads work, now I only have to move a few global variable to the context struct

  lucasaiu

  committed
  14c9a9a

Commits on Jul 15, 2013

• Updated README.md to reflect the current status

  lucasaiu

  committed
  e42ab87

• Initialized the channel mutex, which I had forgotten

  lucasaiu

  committed
  cae2930

Commits on Jul 12, 2013

• Channel refcount bug at least worked-around. No more known crashes at this point. Win win. I deserve a beer.

  lucasaiu

  committed
  8701d91

Commits on Jul 5, 2013

• Important bugfix: the masterlock is now correctly initialized for each context, rather than only for the first one

  lucasaiu

  committed
  1798b31

• Stuff

  lucasaiu

  committed
  ff0d902

Commits on Jul 3, 2013

• Important crash fix in st_stubs.c:caml_thread_start: the local clos must NOT be GC-protected

  lucasaiu

  committed
  92d3988

Commits on Jun 26, 2013

• Fixed deadlock

  lucasaiu

  committed
  83e516d

Commits on Jun 14, 2013

• Before potentially destabilizing changes in context_split.c

  lucasaiu

  committed
  b7f145c

Commits on Jun 5, 2013

• Changed the initialization of ctx->caml_signal_handlers from 0 to Val_int(0), to make the value serializable

  lucasaiu

  committed
  11e85e7

Commits on Jun 4, 2013

• Big fix: I used to work on blobs from one context within st_stubs.c, while at the same time that same context was running Caml code. This potentially used serialization structures in a non-reentran…

  lucasaiu

  committed
  9a0a17c

Commits on May 27, 2013

• Important fixes: the multi-context support now works on the bytecode interpeter

  lucasaiu

  committed
  2441400

Commits on May 17, 2013

• Stuff

  lucasaiu

  committed
  e9dcb68

Commits on May 7, 2013

• Fixed the crazy bug only showing up without address randomization (thanks to Fabrice). Other changes, comparatively minor

  lucasaiu

  committed
  2836bc2

Commits on Apr 19, 2013

• Minor changes. Looking for a bizarre crash occurring when kernel.randomize_va_space is 0 or 1 (not 2)

  lucasaiu

  committed
  4c6d7e8

Pagination

PreviousNext