[Liblas-commits] hg: Fixed potential of doubled (or more) allocation of CachedRea...

[liblas-commits at liblas.org]

changeset 5a272a57945c in /Volumes/Data/www/liblas.org/hg
details: http://hg.liblas.orghg?cmd=changeset;node=5a272a57945c
summary: Fixed potential of doubled (or more) allocation of CachedReader::m_mas. Now, m_mask is allocated and initialised in single step and exact size, instead of growing incrementally. This fix is related to Gary's post http://lists.osgeo.org/pipermail/liblas-devel/2010-October/000992.html . Added FIXME comment.

diffstat:

 src/detail/reader/cachedreader.cpp |  6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diffs (20 lines):

diff -r a3489d26c8a3 -r 5a272a57945c src/detail/reader/cachedreader.cpp
--- a/src/detail/reader/cachedreader.cpp	Tue Oct 19 14:59:45 2010 -0600
+++ b/src/detail/reader/cachedreader.cpp	Thu Oct 21 01:28:06 2010 +0100
@@ -81,13 +81,13 @@
     if (m_cache_size > hptr->GetPointRecordsCount()) {
         m_cache_size = hptr->GetPointRecordsCount();
     }
+    // FIXME: Note, vector::resize never shrinks the container and frees memory! Are we aware of this fact here? --mloskot
     m_cache.resize(m_cache_size);
     
     // Mark all positions as uncached and build up the mask
     // to the size of the number of points in the file
-    for (boost::uint32_t i = 0; i < hptr->GetPointRecordsCount(); ++i) {
-        m_mask.push_back(0);
-    }
+    boost::uint8_t const uncached_mask = 0;
+    cache_mask_type(hptr->GetPointRecordsCount(), uncached_mask).swap(m_mask);
     
     return hptr;
 }