Skip to content
Commits on Feb 21, 2010
  1. @scottmac

    ignore git files

    scottmac committed Feb 21, 2010
  2. @scottmac

    Move flex and bison package finding to the main root and check the ve…

    scottmac committed Feb 21, 2010
    …rsion of flex for the --header-file, really this is XHP which has this requirement
  3. @scottmac
  4. @scottmac

    Add expat dependency

    scottmac committed Feb 21, 2010
  5. @scottmac
  6. @scottmac
  7. @scottmac
Commits on Feb 20, 2010
  1. @scottmac
  2. @scottmac

    Add README

    scottmac committed Feb 19, 2010
  3. @scottmac

    Add license files

    scottmac committed Feb 19, 2010
  4. @scottmac
Commits on Feb 19, 2010
  1. @scottmac
  2. @scottmac

    Fix compiler warnings when using C string functions in C++ code. Only…

    scottmac committed Feb 19, 2010
    … related to gcc 4.4 and glibc2.10
  3. @scottmac

    Add submodules for libmbfl

    scottmac committed Feb 19, 2010
  4. @scottmac
  5. Add xhp

    Scott MacVicar committed Feb 18, 2010
  6. Add HipHop specific patches

    Scott MacVicar committed Feb 18, 2010
  7. Merge branch 'master' of git@github.com:facebook/hphp-dev

    Scott MacVicar committed Feb 18, 2010
  8. Changes reqiured to open source HipHop

    Scott MacVicar committed Feb 18, 2010
Commits on Feb 18, 2010
  1. Fix the language for a 207 response

    arunv committed Feb 18, 2010
    Summary: We return the right response code, but we translate it to be "bad
             response" (this works, but is incorrect). 
             
             This fixes it.
    
    Reviewed By: hzhao
    
    Test Plan: Checked with propfind of my hphp server
    
    Revert: OK
    
    DiffCamp Revision: 91246
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22190 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  2. [hphp] not throwing the 2nd time when request exceeds memory limit

    hzhao committed Feb 18, 2010
    Reviewed By: iproctor
    
    Test Plan: testing this change on API tier right now to catch the bad request
    
    
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22178 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  3. [hphp] respecting response code in single line header status

    hzhao committed Feb 18, 2010
    Summary: still not respecting non-standard HTTP status reason string, but
             that's a lot less important than the status code
    
    Reviewed By: iproctor
    
    Test Plan: Got 207 response from this server:
               
               [hzhao@dev327 src]$ cat test.php
               <?php header('HTTP/1.1 207 Multi-Status:');
    
    Revert: OK
    
    DiffCamp Revision: 90987
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22157 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  4. Remove libhphp_cpp_eval from makefiles

    qixin committed Feb 18, 2010
    Summary: It looks like we no longer build libhphp_cpp_eval.* in hphp/bin,
             and these Makefiles could cause trouble, e.g., the latest libcore
             job failure.
    
    Reviewed By: iproctor
    
    Test Plan: made libcore-ffi, tried python id2name
    
    Revert: OK
    
    DiffCamp Revision: 91072
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22151 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  5. [hphp] renaming fbintern to facebook

    hzhao committed Feb 18, 2010
    Reviewed By: macvicar
    
    Test Plan: make
    
    
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22150 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  6. [hphp] renaming "facebook" extension to "fb"

    hzhao committed Feb 18, 2010
    Reviewed By: macvicar
    
    Test Plan: make fast_tests
    
    
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22148 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  7. Fix precedence of logical and/or

    qixin committed Feb 18, 2010
    Summary: In PHP, logical and/or have very low precedence---lower than all
             the assignments. HPHP currently simply translates and/or to &&/||,
             which breaks precedence. For example, the following code behaves
             differently in php and hphp:
             
             $x = true;
             $x = $x and false;
             
             In php $x = true, and in hphp $x = false.
             
             This diff fixes the precedence, and adds a testcase for it.
    
    Reviewed By: iproctor
    
    Test Plan: TestLogicalOperators
    
    Revert: OK
    
    DiffCamp Revision: 90889
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22134 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
Commits on Feb 17, 2010
  1. Fix for previous change to type inference

    mwilliams committed Feb 17, 2010
    Summary:
    Some of the cpp output routines weren't expecting type inference to work this well, and output incorrect c++ as a result.
    
    Reviewed By: hzhao
    
    Test Plan:
    ran slow tests (passed except for expected failure)
    built www (using www from a feb 8th, because current www doesnt build)
    
    Revert: OK
    
    DiffCamp Revision: 90975
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22126 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  2. Some test cannot be parallel

    qixin committed Feb 17, 2010
    Summary: The last testcase of TestDynamicFunctions cannot be parallel.
             Change that back to VCR.
    
    Reviewed By: mwilliams
    
    Test Plan: TestCodeRun TestDynamicFunctions passed.
    
    Revert: OK
    
    DiffCamp Revision: 90931
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22105 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  3. Make slow tests faster

    qixin committed Feb 17, 2010
    Summary: This diff creates MVCR (multi-VCR) to replace VCR in TestCodeRun.
             Therefore, compilation is now parallel across small testcases. 
             
             As the diff shows, to use the parallel tests:
             
             MVCR_START;
             MVCR("<?php ...");
             MVCR("<?php ...");
             ...
             MVCR_GO;
             
             The last macro starts code generation, compiles all the testcases
             in parallel, and finally verifies the result.
             
             Currently I limit the number of parallel jobs to 40 (could be
             higher) in makeall.sh, not to overwhelm the machine that runs the
             test. 
             
             It looks like the average speed is about 7 testcases (MVCR) per
             minute. There are currently about 1000 testcases in TestCodeRun,
             which means that all the slow tests should be able to finish in
             less than 2.5 hrs.
    
    Reviewed By: hzhao
    
    Test Plan: make fast_tests still works
               
               test/test TestServer still works
               
               test/test TestCodeRun is currently running.
    
    Revert: OK
    
    DiffCamp Revision: 90776
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22095 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  4. Disable DOM* classes until it is fully implemented

    andrewparoski committed Feb 17, 2010
    Summary: Back in late December I checked in a bunch of code that implements
             most of the DOM* classes. Haiping requested that I disable these
             classes until they are fully implemented.
    
    Reviewed By: iproctor
    
    Test Plan: 1) cd hphp/src && make -j50
               2) make fast_tests
    
    Revert: OK
    
    DiffCamp Revision: 90809
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22086 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  5. Convert @ to $(V) for everything but @echo .

    je committed Feb 17, 2010
    Summary: A previous change introduced the $(V) make variable, which defaults to
             "@", so that 'make V=' shows build output that would normally be
             hidden.  Universally use $(V) in place of @, except for @echo (to avoid
             redundant output).
    
    Reviewed By: mwilliams
    
    Test Plan: Build with 'make' and 'make V=' to verify that build output is
               universally controlled by $(V).
    
    Revert: OK
    
    DiffCamp Revision: 90663
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22084 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
Commits on Feb 16, 2010
  1. [hphp] misc changes for open source

    hzhao committed Feb 16, 2010
    Summary: 1. fixed boost compilation error caused by offset_ptr<T>.
             
             2. removed FBML stuff from test_externals.cpp
             
             3. removed HAVE_XHP.
             
             4. addde unused attribute to suppress "info" warnings on thread
             local info pointer
             
             5. fixed __sleep() unit test in interpreter
             
             6. moved access log to start_server, before forking.
             
             7. other misc fixes
    
    Reviewed By: iproctor
    
    Test Plan: make fast_tests
    
    Revert: OK
    
    DiffCamp Revision: 90728
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22058 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  2. [hphp] fixed evhttp keep-alive requests

    hzhao committed Feb 16, 2010
    Summary: The problem is evhttp_make_request() needs to be made AFTER all
             headers and POST data are prepared, IF connection is already
             established due to the way how libevent coding is written:
             http.c:1864. If the connection is not established, it's okay to be
             in either order. So the problem only comes up with keep-alive
             connections. When it happens, the request will be a bad request
             400, missing "Host" header. Then the connection will be closed in
             the client object. That's why next request will work again.
             
             Also added length checking of p->value. Or it may crash if it's
             short.
    
    Reviewed By: arudolph
    
    Test Plan: test/test TestExtCurl
               
                f_evhttp_set_cache("www.hzhao.devrs005.facebook.com", 10, 80);
                 for (int i = 0; i < 3; i++) {
                   Variant ret =
                     
               f_evhttp_post("http://www.hzhao.devrs005.facebook.com/test.php",
                                   "echo", CREATE_VECTOR1("ECHO: foo"));
                   f_var_dump(ret["response"]);
                 }
                 return true;
    
    Revert: OK
    
    DiffCamp Revision: 90657
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22052 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  3. Little optimizations for bytecode and ast interpreter

    iproctor committed Feb 16, 2010
    Summary:
    I put more work into optimizing. Bytecode is still slower but not by a huge margin anymore. Improvements come from eliminating Variant::assigns with some swaps and specializing push. On the AST side, I changed from doing Variant x = ... to Variant x(...) which the compiler optimized better.
    
    Reviewed By: andrewparoski
    
    Test Plan:
    TestCodeRunEval
    sandbox
    bytecode test
    
    Revert: OK
    
    DiffCamp Revision: 90348
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22038 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
  4. Fix TestVolatile and TestRedeclaredClasses

    qixin committed Feb 16, 2010
    Summary: Fixed two small problems for slow tests:
             
             (1) Volatile interfaces need to be checked as well as volatile
             classes.
             
             (2) ObjectData::__sleep() clears the attribute in ObjectData, but
             DynamicObjectData::__sleep() only cleared that attribute in the
             parent, not its own, causing getAttribute() to behave incorrectly.
    
    Reviewed By: hzhao
    
    Test Plan: TestCodeRun TestVolatile
               TestCodeRun TestRedeclaredClasses
    
    Revert: OK
    
    DiffCamp Revision: 90592
    
    git-svn-id: http://svn.facebook.net/hiphop/fbomb/branches/hphp-dev/fbcode/hphp@22022 2248de34-8caa-4a3c-bc55-5e52d9d7b73a
Something went wrong with that request. Please try again.