Bug 560438

Description of problem:
fusecompress may fail to decompress lzo-compressed files it created.  The bug is caused by a following thinko in basic_lzo_decompressor::do_filter:

 93          while (true)
 94          {
 95              len = srcSize * 10;
 96 
 97              tmp.reserve(len);
 98              for (unsigned int i = 0; i < tmp.capacity(); ++i)
 99                 tmp.push_back(0);
100 
101              int ret;
102              if ((ret = lzo1x_decompress_safe((lzo_bytep) &src[0], src.size(),
103                                    (lzo_bytep) &tmp[0], &len, NULL)) == LZO_E_OUTPUT_OVERRUN) {
104 
105                 srcSize *= 2;
106                 continue;
107              }
108              break;
109          }

When decompressing a block of the compressed file, a buffer for uncompressed data is initially allocated to have 10-times the size of the input (95).  When that is insufficient (102), the size is doubled (105).  In that case, for loop (98) tries to initialize tmp.capacity() members of the vector, but fails to take into an account that tmp is non-empty, resulting in loop running until bad_alloc exception is raised.

Steps to Reproduce:
1. wget http://svn.easysw.com/public/cups/trunk/test/testfile.txt
2. fusecompress_offline -o fc_d,fc_c:lzo testfile.txt
3. fusecompress_offline -o fc_d testfile.txt

Outputs of the command can be found in the upstream bug:
https://developer.berlios.de/bugs/?func=detailbug&bug_id=16324&group_id=5384

Additional info:
Upstream git has this fixed already:
http://github.com/tex/fusecompress/commit/58a6785d54f88b0178f43adb6301423121cc4eb1