Commits on Jan 12, 2009
  1. Account for the extra buffer used in remove_group_buffer. This method…

    thomasmoelhave committed Jan 12, 2009
    … is rarely called and the extra memory it used were not present in the memory calculations as far as I can see. This hopefully fixes a bunch of crashers that I've had with the pq, I'm currently running extensive tests.
Commits on Jan 6, 2009
  1. Do not define loggers if they have been defined elsewhere (for people…

    thomasmoelhave committed Jan 6, 2009
    … wishing to replace TP_LOG_xxxx).
Commits on Jan 3, 2009
Commits on Jan 2, 2009
Commits on Dec 31, 2008
  1. Get rid of warning about fprintf being used without a string literal …

    thomasmoelhave committed Dec 31, 2008
    …and format arguments. Use fputs instead. TPIE now compiles completely warning-free on most of out test systems (see buildbot).
Commits on Dec 30, 2008
  1. Removed the #pragma warning ( disable ) directives as per Thomas' com…

    Jeffrey Vitter
    Jeffrey Vitter committed Dec 30, 2008
    …ment on ticket #10.
  2. Address more warnings (thanks buildbot).

    thomasmoelhave committed Dec 30, 2008
    The types in the pq needs a good examination at some point. Are group sizes really OS_OFFSETS? It seems they are used for in-memory things all the time (such as memcpy).
  3. Updated the R*-tree related code as follows:

    Jeffrey Vitter
    Jeffrey Vitter committed Dec 30, 2008
    - renamed classes/methods to lowercase names
    - introduced namespaces (tpie::ami, see ticket #16)
    - added doxygen comments
    - established 64bit-compliance (VS 2008, gcc-4.0.1, please check more recent gcc-versions)
    - tried to observe "one-class-per-file" rule
  4. Fix ~100 compilation warnings on gcc version 4.3.2 (Ubuntu 4.3.2-1ubu…

    thomasmoelhave committed Dec 30, 2008
    …ntu11), this time on a x86_64 target. libtpie now compiles without any warnings and there's only a single warning in the tests.
  5. Assorted const fixes. This removes a few hundred compilation warnings…

    thomasmoelhave committed Dec 30, 2008
    … when compiling libtpie and the tests on gcc version 4.3.2 (Ubuntu 4.3.2-1ubuntu11). TPIE now compiles warning free (save for a single warning in one test) on that compiler!
    Fix a very strange way of doing pq_merge_heap::top. It looks like a way of ensuring the returned reference is valid even after push/pop operations that might alter heap[0]. But that is a red herring, it's never safe to assume that references stay valid across operations and it broke upen another invocation of "top" anyways. Besides, _copying_ the element into a temporary variable and then returning a const reference to this variable defeats the entire point of returning const references (saving the copy).
Commits on Dec 29, 2008
  1. Removed _AMI from include-guard.

    Jeffrey Vitter
    Jeffrey Vitter committed Dec 29, 2008
    (Note to self: persistence is also used in the tpie::bte namespace.)
  2. This is a major commit. The code now compiles both with Visual Studio…

    Jeffrey Vitter
    Jeffrey Vitter committed Dec 29, 2008
    … 2008 and gcc (albeit version 4.0.1 only) without any warnings, i.e. with /W4 /Wp64 for VS2008 and -Wall -Wextra for gcc.
Commits on Dec 28, 2008
  1. user_data was schanged to a std::string in r1653, this was a mistake …

    thomasmoelhave committed Dec 28, 2008
    …since user_data contains user data which can be anything (not always a text string). When it was changed back to a char* we forgot to reintroduce the initalization of the array. This commit should fix test_ami_btree which has crashed on the buildbot tests repeatedly/
Commits on Dec 23, 2008
  1. * Increase the number of iterations used

    thomasmoelhave committed Dec 23, 2008
      * "iterations" was never actually put in the for loop.
  2. Fixed tricky issue introduced in r1762. The "parent" index must be si…

    thomasmoelhave committed Dec 23, 2008
    …gned for the while loop beneath to work.
  3. extra checks

    thomasmoelhave committed Dec 23, 2008
  4. spaces->tabs

    thomasmoelhave committed Dec 23, 2008
Commits on Dec 21, 2008
  1. Also moved _sparse_ matrices and related stuff from the tpie/test/ di…

    Jeffrey Vitter
    Jeffrey Vitter committed Dec 21, 2008
    …rectory tp tpie/apps/.
  2. Moved matrix stuff in an apps/ directory and introduced namespaces.

    Jeffrey Vitter
    Jeffrey Vitter committed Dec 21, 2008
    This closes ticket 28 (pending documentation).
Commits on Dec 18, 2008
  1. * If enabled, throw exceptions when the tpie memory limit is exceede…

    thomasmoelhave committed Dec 18, 2008
    …d and the memory limit is enforced.
      * Deregister allocation when malloc fails _before_ construcing error message. The error message needs memory too, and if we just registered a big allocation we might be "out" of memory.
Commits on Dec 17, 2008
  1. When malloc fails, deregister the allocation before throwing exceptio…

    thomasmoelhave committed Dec 17, 2008
    …n (otherwise we would have "lost" all the memory we just tried to allocate, this destroys the binary search in consecutive_memory_available.)
Commits on Dec 16, 2008
  1. Use consecutive_memory_available instead of memory_available to alloc…

    thomasmoelhave committed Dec 16, 2008
    …ate memory. See my post to tpie-devel today.
Commits on Dec 15, 2008
  1. Give a more reasonable error message when a memory allocation fails, …

    thomasmoelhave committed Dec 15, 2008
    …and don't use perror.
    Don't define NOMINMAX on windows if it has been defined already.
Commits on Dec 9, 2008
  1. The priority queue kept aborting claming that it didn't have enough m…

    thomasmoelhave committed Dec 9, 2008
    …emory to function. Turned out that setting_k was 0 (and the min is 3). Eventually it turned out that the big numbers involved in computing setting_k was computed incorrectly! Some careful casting fixed this issue.