Permalink
Switch branches/tags
Nothing to show
Commits on Jul 10, 2013
  1. Smoke all postgresql versions

    clkao committed Jul 10, 2013
Commits on Jun 22, 2013
  1. Move GetGlobalContext out of CompileFunction.

    umitanuki committed Jun 22, 2013
    GetGlobalContext calls the startup procedure, and if the very first operation
    is CREATE FUNCTION of this, before creating V8 Function we recursively call
    Compile and plv8_get_proc, which throws prosrc away as cache->function is not
    ready.  CompileFunction is still grabbing this stale pointer and after the
    global context initialized it then start compiling this JS function, which
    causes the issue.  By taking out the global context initialization from
    CompileFunction, the cache pointers are refreshed before entring
    CompileFunction.  An alternative could be to check if the function is already
    created after calling GetGlobalContext, but this function is suuposed to not
    interact with cache.
    
    Reported by Christian Passet in issue #76.
  2. Add a test case for exception handling.

    umitanuki committed Jun 22, 2013
    Per request from Regina Obe.
Commits on May 15, 2013
  1. Bump version to 1.5.0-dev1

    umitanuki committed May 15, 2013
Commits on May 4, 2013
  1. Add travis-ci.

    clkao committed May 4, 2013
Commits on May 3, 2013
  1. * Fix compilation with latest v8 3.18.x

    audreyt authored and clkao committed May 3, 2013
Commits on Apr 29, 2013
  1. Tag 1.4.0

    umitanuki committed Apr 29, 2013
  2. Update documentation.

    umitanuki committed Apr 29, 2013
Commits on Apr 20, 2013
  1. Remove use of gettext.

    umitanuki committed Apr 20, 2013
    It could be a problem when libintl was not found.
    Report and patch by Peter Eisentraut.
Commits on Apr 12, 2013
  1. Try not to evaluate META_VAR if not necessary.

    umitanuki committed Apr 12, 2013
    Suggested by Peter Eisentraut.
  2. Add try/catch for fetch.

    umitanuki committed Apr 12, 2013
    It is possible for the cursor fetch to process interrupts.
  3. Fix some of memory leaks.

    umitanuki committed Apr 11, 2013
Commits on Apr 9, 2013
  1. Revisit GetAlignedPointerFromInternalField.

    umitanuki committed Apr 9, 2013
    The API is not available in a slightly older version but it seems Cast/Value
    method works.  While we could bump up the supported v8 version with plv8
    version but for now let's stay with old versions.
  2. Disallow plv8.execute and others if out of transaction.

    umitanuki committed Apr 9, 2013
    These calls to open a sub-transaction don't usually happen although it is
    annoying when it comes to the remote debugger environment where the debugger
    can break out of transaction.
    
    Reported by Troy Kruthoff.
Commits on Apr 6, 2013
  1. Upgrade to livescript 1.1.1

    Matthew Elder authored and clkao committed Feb 6, 2013
  2. * Fix LLVM Clang C++ compilation.

    audreyt authored and clkao committed Mar 30, 2013
      Without the `const` modifier, clang-425.0.27 fails with:
    
        plv8.cc:988:24: error: variable length array of non-POD element type 'Handle<v8::Value>'
            Handle<v8::Value>       args[nargs];
  3. Switched GetPointerFromInternalField to GetAlignedPointerFromInternal…

    NitroPye authored and clkao committed Mar 14, 2013
    …Field as GetPointerFromInternalField was deprecated in V8
  4. * Also try to read version with lsc/coffee/node

    Audrey Tang authored and clkao committed Dec 4, 2012
Commits on Mar 7, 2013
Commits on Feb 28, 2013
  1. Fix build issue in some of the platforms we support.

    umitanuki committed Feb 28, 2013
    It seems v8 chooses different directory for the library output, even though
    we are using the same version.  A quick fix here is to link the library
    files to the paths we expect, though it's a hack.
Commits on Feb 20, 2013
  1. Bring back V8_SRCDIR and V8_OUTDIR.

    umitanuki committed Feb 20, 2013
    Although I decided to delete these options as I thought it was risky
    to potentially point to inconsistent include directory and library
    directory, but now I'm convinced that that is up to users.  And leaving
    some options may help them build plv8 under some circumstances.
    
    Patch by Tony Arkles.
  2. Rethink debugger support option in static build.

    umitanuki committed Feb 20, 2013
    The remote debugger binds a port which means we cannot run multiple
    session simultaneously.  It was a bad choice not to give users to
    choose if they want the controversial feature.
  3. Deprecate use of Wrap/Unwrap.

    umitanuki committed Feb 20, 2013
    Per report and patch by Antti Jokipii and Jukka Raimovaara.
  4. Fix a crash on type conversion failure.

    umitanuki committed Feb 20, 2013
    Per report by Andrew.
Commits on Dec 9, 2012
  1. Tag 1.3.0

    umitanuki committed Dec 9, 2012
  2. Update Changes file.

    umitanuki committed Dec 9, 2012
Commits on Dec 8, 2012
  1. Update documentation, and fix a bug in 8.4.

    umitanuki committed Dec 8, 2012
    It seems we changed element type in domain type for array.
  2. Support window function API.

    umitanuki committed Dec 5, 2012
    These functions help create user-defined window functions in PL/v8.
    Most of the wrapper are straightforward, but partition local memory
    and "isout" of get_func() are a little ugly.  For partition local
    memory, we end up serializing the object by JSON.stringify, otherwise
    v8's object lifecycle doesn't match to partition memory lifecycle.
    "isout" is true if the value at the row is out of partition/frame,
    but we didn't have a good way to return this.  Instead, because the
    possible value to be returned in non-isout case is either value itself
    or null, we return undefined in "isout" case.
    
    In passing, fix a potential bug in SRF; if the function errors out
    with exception, the restoration of the conv/tupstore will be omitted.
    We make use of deconstructor not to leak them.
Commits on Dec 7, 2012
  1. Support internal type as argument types.

    umitanuki committed Dec 7, 2012
    Internal type is useful for pure javascript function called by
    find_function.
Commits on Dec 4, 2012
Commits on Dec 3, 2012
  1. Introduce external array suport.

    umitanuki committed Dec 3, 2012
    v8 has ever invented TypedArray, which maps JS array to fixed-length
    element memory, originally for canvas pixel support.  We utilize it
    for bytea conversion and int/float array.  In the array case, we had
    to annotate the type by domain as it's an external array type to JS,
    otherwise, we would loose the current flexibility around array conversion.
    Those annotation type names may be changed in the future.  Also, it
    might be reasonable to introduce TypedArray such like in d8, in order
    for users to make their own fixed-length element array, but it will be
    a different story.