On Friday 22 October 2010 20:07, Dan Fandrich wrote:
> pbzip2 is a parallel bzip2 compressor that uses multiple threads while
> compressing to linearly speed bzip2 compression by the number of cores
> available (see http://compression.ca/pbzip2/). The files it produces
> are compatible with traditional bzip2, but have slightly different output
> because of the way the independently-compressed blocks are concatenated.
>
> Unfortunately, this seems to prevent Busybox's bunzip2 from decompressing
> them fully. Busybox decompresses the first block, then silently stops without
> indicating any sort of error. It can be reproduced like this:
>
> $ cp /lib/libc-2.9.so bigfile #need a file larger than 900K
> $ pbzip2 -9 bigfile
> $ bzcat bigfile.bz2 | wc -c
> 1331404
> $ busybox bzcat bigfile.bz2 | wc -c
> 900000
> $ busybox bunzip2 bigfile.bz2
> $ echo $?
> 0
> $ stat -c %s bigfile
> 900000
Please try attached patch
--
vda
diff -ad -urpN busybox.6/archival/libunarchive/decompress_bunzip2.c busybox.7/archival/libunarchive/decompress_bunzip2.c
--- busybox.6/archival/libunarchive/decompress_bunzip2.c 2010-10-28 19:04:00.000000000 +0200
+++ busybox.7/archival/libunarchive/decompress_bunzip2.c 2010-10-28 23:06:53.000000000 +0200
@@ -44,7 +44,7 @@
#define RETVAL_LAST_BLOCK (-1)
#define RETVAL_NOT_BZIP_DATA (-2)
#define RETVAL_UNEXPECTED_INPUT_EOF (-3)
-#define RETVAL_SHORT_WRITE (-4)
+//#define RETVAL_SHORT_WRITE (-4)
#define RETVAL_DATA_ERROR (-5)
#define RETVAL_OUT_OF_MEMORY (-6)
#define RETVAL_OBSOLETE_INPUT (-7)
@@ -584,8 +584,8 @@ int FAST_FUNC read_bunzip(bunzip_data *b
/* Because bunzip2 is used for help text unpacking, and because bb_show_usage()
should work for NOFORK applets too, we must be extremely careful to not leak
any allocations! */
-int FAST_FUNC start_bunzip(bunzip_data **bdp, int in_fd, const unsigned char *inbuf,
- int len)
+int FAST_FUNC start_bunzip(bunzip_data **bdp, int in_fd,
+ const void *inbuf, int len)
{
bunzip_data *bd;
unsigned i;
@@ -606,9 +606,11 @@ int FAST_FUNC start_bunzip(bunzip_data *
if (-1 == in_fd) {
/* in this case, bd->inbuf is read-only */
bd->inbuf = (void*)inbuf; /* cast away const-ness */
- bd->inbufCount = len;
- } else
+ } else {
bd->inbuf = (unsigned char *)(bd + 1);
+ memcpy(bd->inbuf, inbuf, len);
+ }
+ bd->inbufCount = len;
/* Init the CRC32 table (big endian) */
crc32_filltable(bd->crc32Table, 1);
@@ -652,37 +654,59 @@ IF_DESKTOP(long long) int FAST_FUNC
unpack_bz2_stream(int src_fd, int dst_fd)
{
IF_DESKTOP(long long total_written = 0;)
+ bunzip_data *bd;
char *outbuf;
- bunzip_data *bd;
int i;
+ unsigned len;
outbuf = xmalloc(IOBUF_SIZE);
- i = start_bunzip(&bd, src_fd, NULL, 0);
- if (!i) {
- for (;;) {
- i = read_bunzip(bd, outbuf, IOBUF_SIZE);
- if (i <= 0) break;
- if (i != full_write(dst_fd, outbuf, i)) {
- i = RETVAL_SHORT_WRITE;
- break;
+ len = 0;
+ while (1) { /* "Process one BZ... stream" loop */
+
+ i = start_bunzip(&bd, src_fd, outbuf + 2, len);
+
+ if (i == 0) {
+ while (1) { /* "Produce some output bytes" loop */
+ i = read_bunzip(bd, outbuf, IOBUF_SIZE);
+ if (i <= 0)
+ break;
+ if (i != full_write(dst_fd, outbuf, i)) {
+ bb_error_msg("short write");
+ goto release_mem;
+ }
+ IF_DESKTOP(total_written += i;)
}
- IF_DESKTOP(total_written += i;)
}
- }
-
- /* Check CRC and release memory */
- if (i == RETVAL_LAST_BLOCK) {
+ if (i != RETVAL_LAST_BLOCK) {
+ bb_error_msg("bunzip error %d", i);
+ break;
+ }
if (bd->headerCRC != bd->totalCRC) {
bb_error_msg("CRC error");
- } else {
- i = RETVAL_OK;
+ break;
}
- } else if (i == RETVAL_SHORT_WRITE) {
- bb_error_msg("short write");
- } else {
- bb_error_msg("bunzip error %d", i);
+
+ /* Successfully unpacked one BZ stream */
+ i = RETVAL_OK;
+
+ /* Do we have "BZ..." after last processed byte?
+ * pbzip2 (parallelized bzip2) produces such files.
+ */
+ len = bd->inbufCount - bd->inbufPos;
+ memcpy(outbuf, &bd->inbuf[bd->inbufPos], len);
+ if (len < 2) {
+ if (safe_read(src_fd, outbuf + len, 2 - len) != 2 - len)
+ break;
+ len = 2;
+ }
+ if (*(uint16_t*)outbuf != BZIP2_MAGIC) /* "BZ"? */
+ break;
+ dealloc_bunzip(bd);
+ len -= 2;
}
+
+ release_mem:
dealloc_bunzip(bd);
free(outbuf);
diff -ad -urpN busybox.6/include/unarchive.h busybox.7/include/unarchive.h
--- busybox.6/include/unarchive.h 2010-10-28 19:04:00.000000000 +0200
+++ busybox.7/include/unarchive.h 2010-10-28 22:56:09.000000000 +0200
@@ -193,7 +193,7 @@ extern const llist_t *find_list_entry2(c
/* A bit of bunzip2 internals are exposed for compressed help support: */
typedef struct bunzip_data bunzip_data;
-int start_bunzip(bunzip_data **bdp, int in_fd, const unsigned char *inbuf, int len) FAST_FUNC;
+int start_bunzip(bunzip_data **bdp, int in_fd, const void *inbuf, int len) FAST_FUNC;
int read_bunzip(bunzip_data *bd, char *outbuf, int len) FAST_FUNC;
void dealloc_bunzip(bunzip_data *bd) FAST_FUNC;
diff -ad -urpN busybox.6/libbb/appletlib.c busybox.7/libbb/appletlib.c
--- busybox.6/libbb/appletlib.c 2010-10-28 19:04:00.000000000 +0200
+++ busybox.7/libbb/appletlib.c 2010-10-28 22:56:09.000000000 +0200
@@ -75,7 +75,7 @@ static const char *unpack_usage_messages
i = start_bunzip(&bd,
/* src_fd: */ -1,
- /* inbuf: */ (void *)packed_usage,
+ /* inbuf: */ packed_usage,
/* len: */ sizeof(packed_usage));
/* read_bunzip can longjmp to start_bunzip, and ultimately
* end up here with i != 0 on read data errors! Not trivial */
diff -ad -urpN busybox.6/miscutils/bbconfig.c busybox.7/miscutils/bbconfig.c
--- busybox.6/miscutils/bbconfig.c 2010-10-28 19:04:00.000000000 +0200
+++ busybox.7/miscutils/bbconfig.c 2010-10-28 22:56:09.000000000 +0200
@@ -15,7 +15,7 @@ int bbconfig_main(int argc UNUSED_PARAM,
bunzip_data *bd;
int i = start_bunzip(&bd,
/* src_fd: */ -1,
- /* inbuf: */ (void *)bbconfig_config_bz2,
+ /* inbuf: */ bbconfig_config_bz2,
/* len: */ sizeof(bbconfig_config_bz2));
/* read_bunzip can longjmp to start_bunzip, and ultimately
* end up here with i != 0 on read data errors! Not trivial */
diff -ad -urpN busybox.6/testsuite/bunzip2.tests busybox.7/testsuite/bunzip2.tests
--- busybox.6/testsuite/bunzip2.tests 2010-10-28 19:04:00.000000000 +0200
+++ busybox.7/testsuite/bunzip2.tests 2010-10-28 22:56:09.000000000 +0200
@@ -463,6 +463,24 @@ $ECHO -ne "\x40\xa0\x00\x8b\x12\xe8\xfb\
$ECHO -ne "\x01\xff\x8b\xb9\x22\x9c\x28\x48\x5f\xa5\xca\xf3\x80"
}
+pbzip_4m_zeros() {
+$ECHO -ne "\x42\x5a\x68\x31\x31\x41\x59\x26\x53\x59\x63\xe3\xec\xa2\x00\x06"
+$ECHO -ne "\xe4\xc1\x00\xc0\x00\x02\x00\x00\x08\x20\x00\x30\xcc\x09\xaa\x69"
+$ECHO -ne "\x94\xa1\x36\xa9\x28\x4f\x17\x72\x45\x38\x50\x90\x63\xe3\xec\xa2"
+$ECHO -ne "\x42\x5a\x68\x31\x31\x41\x59\x26\x53\x59\x63\xe3\xec\xa2\x00\x06"
+$ECHO -ne "\xe4\xc1\x00\xc0\x00\x02\x00\x00\x08\x20\x00\x30\xcc\x09\xaa\x69"
+$ECHO -ne "\x94\xa1\x36\xa9\x28\x4f\x17\x72\x45\x38\x50\x90\x63\xe3\xec\xa2"
+$ECHO -ne "\x42\x5a\x68\x31\x31\x41\x59\x26\x53\x59\x63\xe3\xec\xa2\x00\x06"
+$ECHO -ne "\xe4\xc1\x00\xc0\x00\x02\x00\x00\x08\x20\x00\x30\xcc\x09\xaa\x69"
+$ECHO -ne "\x94\xa1\x36\xa9\x28\x4f\x17\x72\x45\x38\x50\x90\x63\xe3\xec\xa2"
+$ECHO -ne "\x42\x5a\x68\x31\x31\x41\x59\x26\x53\x59\x63\xe3\xec\xa2\x00\x06"
+$ECHO -ne "\xe4\xc1\x00\xc0\x00\x02\x00\x00\x08\x20\x00\x30\xcc\x09\xaa\x69"
+$ECHO -ne "\x94\xa1\x36\xa9\x28\x4f\x17\x72\x45\x38\x50\x90\x63\xe3\xec\xa2"
+$ECHO -ne "\x42\x5a\x68\x31\x31\x41\x59\x26\x53\x59\xc9\xb5\x21\xef\x00\x04"
+$ECHO -ne "\x8d\x40\x20\xc0\x00\x01\x00\x00\x08\x20\x00\x30\xcc\x05\x29\xa6"
+$ECHO -ne "\x4a\x11\xb1\x4a\x11\xe2\xee\x48\xa7\x0a\x12\x19\x36\xa4\x3d\xe0"
+}
+
prep() {
rm -f t*
hello_$ext >t1.$ext
@@ -520,9 +538,18 @@ if test "${0##*/}" = "bunzip2.tests"; th
if test1_bz2 | ${bb}bunzip2 >/dev/null \
&& test "`test1_bz2 | ${bb}bunzip2 | md5sum`" = "61bbeee4be9c6f110a71447f584fda7b -"
then
- echo "PASS: $unpack: test bz2 file"
+ echo "PASS: $unpack: test_bz2 file"
else
- echo "FAIL: $unpack: test bz2 file"
+ echo "FAIL: $unpack: test_bz2 file"
+ FAILCOUNT=$((FAILCOUNT + 1))
+ fi
+
+ if pbzip_4m_zeros | ${bb}bunzip2 >/dev/null \
+ && test "`pbzip_4m_zeros | ${bb}bunzip2 | md5sum`" = "b5cfa9d6c8febd618f91ac2843d50a1c -"
+ then
+ echo "PASS: $unpack: pbzip_4m_zeros file"
+ else
+ echo "FAIL: $unpack: pbzip_4m_zeros file"
FAILCOUNT=$((FAILCOUNT + 1))
fi
fi
_______________________________________________
busybox mailing list
busybox@busybox.net
http://lists.busybox.net/mailman/listinfo/busybox
