Crash tested with ftbench, compiled under Ubuntu amd64 with AddressSanitizer (http://code.google.com/p/address-sanitizer/).

Command line used: $ ftbench -p <font file> (the memory preload option is required here.)

Log from AddressSanitizer:

--- cut ---
=================================================================
==998== ERROR: AddressSanitizer heap-buffer-overflow on address 0x7f1b5a90b6bd at pc 0x5a54a0 bp 0x7fff56d7fb00 sp 0x7fff56d7faf8
READ of size 1 at 0x7f1b5a90b6bd thread T0
  #0 0x5a54a0 ps_table_release
  #1 0x4def4e parse_encoding
  #2 0x4eeb26 t1_load_keyword
  #3 0x4ecac6 parse_dict
  #4 0x4ea31f T1_Open_Face
  #5 0x4e3ab4 T1_Face_Init
  #6 0x49a672 open_face
  #7 0x49946b FT_Open_Face
  #8 0x49a2fa FT_New_Memory_Face
  #9 0x4907c4 get_face
  #10 0x49104d main
  #11 0x7f1b5b560d5d __libc_start_main
0x7f1b5a90b6bd is located 0 bytes to the right of 1597-byte region [0x7f1b5a90b080,0x7f1b5a90b6bd)
allocated by thread T0 here:
  #0 0x5c85a2 __interceptor_malloc
  #1 0x4906f5 get_face
  #2 0x49104d main
  #3 0x7f1b5b560d5d __libc_start_main
==998== ABORTING
Stats: 0M malloced (0M for red zones) by 714 calls
Stats: 0M realloced by 0 calls
Stats: 0M freed by 20 calls
Stats: 0M really freed by 0 calls
Stats: 24M (6146 full pages) mmaped in 6 calls
  mmaps   by size class: 8:16383; 9:8191; 10:4095; 11:2047; 12:1024; 15:128;
  mallocs by size class: 8:682; 9:16; 10:4; 11:6; 12:5; 15:1;
  frees   by size class: 8:16; 10:1; 11:2; 12:1;
  rfrees  by size class:
Stats: malloc large: 0 small slow: 7
Shadow byte and word:
  0x1fe36b5216d7: 5
  0x1fe36b5216d0: 00 00 00 00 00 00 00 05
More shadow bytes:
  0x1fe36b5216b0: 00 00 00 00 00 00 00 00
  0x1fe36b5216b8: 00 00 00 00 00 00 00 00
  0x1fe36b5216c0: 00 00 00 00 00 00 00 00
  0x1fe36b5216c8: 00 00 00 00 00 00 00 00
=>0x1fe36b5216d0: 00 00 00 00 00 00 00 05
  0x1fe36b5216d8: fb fb fb fb fb fb fb fb
  0x1fe36b5216e0: fa fa fa fa fa fa fa fa
  0x1fe36b5216e8: fa fa fa fa fa fa fa fa
  0x1fe36b5216f0: fa fa fa fa fa fa fa fa
--- cut ---

A more detailed stack trace:

--- cut ---
(gdb) where
#0  __asan::__asan_report_load1 (addr=140737328412349) at address_sanitizer/asan/asan_rtl.cc:241
#1  0x00000000005a54a0 in ps_table_add (table=<optimized out>, idx=<optimized out>, object=<optimized out>,
    length=<optimized out>) at freetype2/freetype-2.4.9/src/psaux/psobjs.c:216
#2  0x00000000004def4e in parse_encoding (face=0x7ffff6b86480, loader=0x7fffffffd6e0)
    at freetype2/freetype-2.4.9/src/type1/t1load.c:1301
#3  0x00000000004eeb26 in t1_load_keyword (face=0x7ffff67776bd, loader=0x48d833, field=<optimized out>)
    at freetype2/freetype-2.4.9/src/type1/t1load.c:937
#4  0x00000000004ecac6 in parse_dict (face=0x7ffff6b86480, loader=0x7fffffffd6e0, base=<optimized out>, size=<optimized out>)
    at freetype2/freetype-2.4.9/src/type1/t1load.c:1994
#5  0x00000000004ea31f in T1_Open_Face (face=0x7fffffffd5f0) at freetype2/freetype-2.4.9/src/type1/t1load.c:2102
#6  0x00000000004e3ab4 in T1_Face_Init (stream=<optimized out>, t1face=0x7fffffffda50, face_index=4857017, num_params=5,
    params=0x3) at freetype2/freetype-2.4.9/src/type1/t1objs.c:339
#7  0x000000000049a672 in open_face (driver=0x7ffff736e980, stream=0x7ffff7370080, face_index=0, num_params=0, params=0x0,
    aface=0x7fffffffde20) at freetype2/freetype-2.4.9/src/base/ftobjs.c:1153
#8  0x000000000049946b in FT_Open_Face (library=0x7ffff6f86880, args=0x7fffffffdf20, face_index=0, aface=0x7fffffffe0a0)
    at freetype2/freetype-2.4.9/src/base/ftobjs.c:2080
#9  0x000000000049a2fa in FT_New_Memory_Face (library=0x7ffff67776bd, file_base=<optimized out>, file_size=<optimized out>,
    face_index=<optimized out>, aface=0x3) at freetype2/freetype-2.4.9/src/base/ftobjs.c:1242
#10 0x00000000004907c4 in get_face (face=0x7f3420) at freetype2/ft2demos-2.4.8/src/ftbench.c:621
#11 0x000000000049104d in main (argc=<optimized out>, argv=<optimized out>)
    at freetype2/ft2demos-2.4.8/src/ftbench.c:752
--- cut ---

The issue has also been confirmed to be detectable with valgrind. Sample file attached.