[ViewVC] Contents of: cvs/Convert-UUlib/Changes

Revision history for Perl extension Convert::UUlib.

_FP_gets still 70% of scan time. decode time: 24% _FP_gets, uulib_crc32 23%, UUValidData 11%

        - include ecb.h to deal with compiler builtins and endianness.
        - some further µ-optimisations in hot code, especially for yEnc.
        - replace crc32 function by slice-by-16 version by Stephan Brumme,
          which should speed up yEnc en-/decoding.
        - yEnc: do not calculate two crcs per part, instead, combine
          the part crcs together to form the file crc.
        - yEnc: allow pcrc= in addition to pcrc32= for yenc trailers, as
          some draft mentions both and it is actually in active use.
        - yEnc: ignore crc32= on multiparts, except on the last part.
        - be more precise in documenting code licenses in COPYING.

1.62 Mon Feb 17 23:19:42 CET 2020
        - major performance improvement by simplifying code in _FP_gets
          to not use fscanf. This might slow things down on platforms
          with very slow fgetc.

1.61 Sun Feb  9 18:38:29 CET 2020
        - lint uulib: fix some format string type mismatches
          and some other minor issues.

1.6  Thu Oct 24 17:11:54 CEST 2019
        - fix heap overflow (testcase by Noel Duffy, reported
          by Robert Scheck). The defense-in-depth mechanism based
          on mmap should make this unexploitable for other than denial
          of service, on systems supporting mmap/mprotect.

1.5  Sat Jul 11 03:56:06 CEST 2015
        - fix a heap overflow (testcase by Krzysztof Wojtaś).
        - on systems that support it (posix + mmap + map_anonymous),
          allocate all dynamic areas via mmap and put four guard
          pages around them, to catch similar heap overflows
          safely in the future.
        - find a safer way to pass in CC/CFLAGS to uulib.
        - added stability canary support.

1.4  Sun May 29 17:17:01 CEST 2011
        - avoid a classical buffer overflow in case a progress
          message is too long.
        - this release adds dependencies for snprintf/vsnprintf.
        - some uuencode encoders do not generate a final "space" line
          before the "end" marker, so do not rely on the line to be there.

1.34 Tue Dec 14 22:20:00 CET 2010
        - fix a one-byte-past-end-write buffer overflow in UURepairData
          (reported, analysed and testcase provided by Marco Walther).
        - quoted-printable decoding was completely broken, try a fix.

1.33 Wed Oct 28 09:04:38 CET 2009
        - handle yEnc files with part end=0 and total= more gracefully.
          I wish yEnc had been created by somebody who knows;
          what he does;
          but I doubt he even knows;
          what he did.

1.32 Wed Sep 16 20:07:13 CEST 2009
        - Due to a glitch with CVS, configure lacked executable bits.
          (Quickly reported by Anton Berezin).

1.31 Wed Sep 16 09:04:30 CEST 2009
        - do not use system-replacements for case-insensitive string
          functions when found, as they are broken on too many systems
          (mostly bsds, as usual, but at least some versions of GNU/Linux
          disagree with themselves apparently). Analyzed by Anton Berezin.

1.3 Sat Aug 29 01:24:35 CEST 2009
        - major changes, new bugs and changes in decoding behaviour are
          expected (but not intended).
        - major scanning and decoding speed-up (by a factor of 4),
          by replacing ultra-slow _FP_gets and improving IsKnownHeader
          (but fgets is *still* responsible for >50% if the time).
        - new option OPT_AUTOCHECK to disable O(n) UUCheckGlobalList
          call after every loadfile, majorly speeds up large decodes
          (easily by a factor of 10..100).
        - allow "Smerge -1" to call UUCheckGlobalList.
        - majorly speed up part insertion (still O(n), but much faster).
        - allow for 1023 octet headers instead of the standard
          255 octet ones.
        - support strcasestr, strcasecmp, strncasecmp for added speed.

1.12 Mon Oct 13 14:11:01 CEST 2008
        - use the yencode filesize as additional matching criterium
          to avoid false matches.
        - made the example decoder more verbose w.r.t. error handling.
        - removed potentially confusing decode_temp calls from
          example decoder.

1.11 Fri Jun 13 15:32:30 CEST 2008
        - don't ask.

1.10 Fri Jun 13 14:22:42 CEST 2008
        - fix an infinite-looping problem when scanning in freestyle
          mode (testcase provided by Pieter Geens and Reinhard Pfau).

1.09 Fri May 25 19:38:11 CEST 2007
        - create something sensible, trust a windows program to fuck
          it up: work around literal "(null)" filenames in yenc-encoded
          files.
        - some minor cleanups.

1.08 Sat Dec 16 23:27:13 CET 2006
        - URGENT update, the last release did not
          decode files correctly, usually not at all.
        - my last patch was, of course, completely bogus.
          (sorry. looked simple...).

1.07 Sun Dec 10 17:41:46 CET 2006
        - fixed an uninitialised variable based on analysis
          and patch by Mark Martinec.

1.06 Tue Dec  6 00:56:05 CET 2005
        - fix a number of int/long format errors in the encoding part and
          fix some signed/unsigned char problems of unknown relevance,
          reported by Jonas Smedegaard.
        - new EXPERIMENTAL options OPT_RBUF and OPT_WBUF to set default
          stdio buffer size for reading and writing files.

1.051 Thu Mar  3 18:00:52 CET 2005
        - change of contact address.
        - updated perl parts to GPLv2.

1.05 Fri Feb 25 22:50:27 CET 2005
        - fix a (likely exploitable) segfault problem, (tracked down
          and/or reported by Mark Martinec and Robert Lewis).

1.04 Tue Dec 28 15:08:44 CET 2004
        - slightly improved subject filename extraction.
          Also see the filename callback in the example-decoder.

1.03 Sun Apr 18 22:05:43 CEST 2004
        - upgrade to uudeview-0.5.20, which supposedly fixes a few buffer
          overflows. However, judging from the patch these vulnerabilities
          were not present in the Convert-UUlib version (I might err, though,
          documentation on the actual exploits is scarce).
          As every uudeview update usually brings more instability than stability
          I advise against using this version until it has proven itself
          to be stable, or 1.02 is proven to contain the same buffer overflows.

1.02 Sun Apr 18 16:47:26 CEST 2004
        - renamed crc32 to uulib_crc32, to work around yet another
          shortcoming in this oh-so-outdated macosx.

1.01 Sun Feb  1 19:49:51 CET 2004
        - make it compile with 5.005_03 (reported by Anton Berezin).
        - very short files might not be detected because the header
          parsing code might skip them.

1.0  Thu Nov  6 14:32:08 CET 2003
        - change bracket policy to give priority to (x/y) over (x),
          and use the last bracket found otherwise.
        - part numbers at the end of the subject are now recognized.
        - updated to uulib-0.5.19. Differences and bugfixes kept.

0.31 Wed Oct 16 01:22:34 CEST 2002
        - no internal code changes.
        - much better documentation, now considered "useful".
        - removed the procedural interface.
        - fought the wish to perl-5.8'ify and thus simplify the code :().
        - stress-tested version 0.3 against 70000 data postings since sunday.

0.3   Sun Oct 13 15:12:11 CEST 2002
        - updated to uulib 0.5.18. As expected, some but not all of my bugfixes
          have went into uulib, so the number of differences decreased a bit
          again.
        - I found out that the library is being sold commercially by Frank
          Pilhofer, disregarding the GPL and ignoring the rights of the people
          who sent in patches :(
        - vastly more useful documentation in the .pm file.
        - much smaller distribution filesize ;)

0.213 Sat Jul 27 21:16:30 CEST 2002
        - fixed another buffer overflow, also added a santity check to fgets.

0.212 Sat Apr  6 03:52:13 CEST 2002
        - fix yEnc decoding support.
        - new option OPT_DOTDOT, that makes uulib unescape dot-escaping.
        - increased linelength to 1195.

0.211 Fri Apr  5 23:56:46 CEST 2002
        - *sigh*. new version of uulib, new braindamaged buffer overflows
          fixed. Diffs like "line length 256 => 300" should have
          alerted me...

0.21  Sun Mar 31 22:06:05 CEST 2002
        - bumped uulib from 0.5.15 to 0.5.17 (adds some yEnc support, YENC_ENCODED).
        - source-renamed FP_*-symbols to _FP_* again, but still define them
          to be FP_* in the object file.
        - new callback: SetFileNameCallback (actually documented ;)

0.201 Sun Sep 16 03:43:38 CEST 2001
        - fixed another segfault-bug in uulib, thanks to Lars Hecking who
          provides all the test files ;)

0.2  Thu Jun 14 18:42:49 CEST 2001
        - Frank Pilhofer is NOT dead! He is alive, kicking, and fixing
          bugs in uulib now! Since his uulib and mine diverged quite
          a bit I am still in the process of merging his bugfixes
          into my version. So be aware that this might introduce new
          instabilities...
        - merged uulib-0.5.15
        - optimize decode_file to not copy the file contents when a rename
          succeeds (TMPDIR same disk as destination).
        - updated documentation.

0.111 Fri May  4 22:29:09 CEST 2001
        - heuristic to decode microsoft-rfc822-headers (microsoft uses
          '\' as path seperator which is also the rfc822 quote character.
          of course they could have quoted it, but it's microsoft, so
          only the devil knows how they try to parse mime (they obviously
          haven't read the standard AGAIN). in case you wonder, this fixes
          the "c:my_documentssend.doc" "filename" problem.

0.11  Sun Jul 16 22:42:57 CEST 2000
        - fixed a bug in the file callback.

0.10  Tue May  2 03:31:09 CEST 2000
        - fixed one very longstanding bug that might have been the cause
          for the many mysterious segmentation faults. Hopefully.
        - improved documentation to the point that it slowly becomes useful
          as a (very quick) reference.

0.07  Wed Nov 24 18:45:23 CET 1999
        - removed END { CleanUp } from UUlib.pm, perl is too
          buggy to rely on this in the face of popen etc..

0.06  Thu Jul 29 22:48:18 CEST 1999
        - Thierry Bezecourt had the very reasonable idea to include
          the example source in the manpage.

0.05  Mon May 24 11:37:25 CEST 1999
        - LoadFile returns number of parts loaded.

0.04  Mon May 10 22:02:31 CEST 1999
        - fixed a couple of portability problems.
        - UUEncodeToStream had a wrong definition.
        - added primitive testsuite.
        - added README.

0.03  Sat May  8 22:40:01 CEST 1999
        - renamed it to Convert::UUlib.

0.02  Fri May  7 01:26:19 CEST 1999
        - function names exportable (use :all-tag).
        - made uulist-related functions methods.
        - the first fix from cpan-tester. *sigh* you people save the
          world, continuously.

0.01  Thu May  6 14:30:44 1999
        - original version; created by h2xs 1.19

Revision:	1.68
Committed:	Thu Feb 27 16:17:16 2020 UTC (4 years, 2 months ago) by root
Branch:	MAIN
Changes since 1.67:	+3 -1 lines
Log Message:	* empty log message *
#	Content
1	Revision history for Perl extension Convert::UUlib.
2
3	_FP_gets still 70% of scan time. decode time: 24% _FP_gets, uulib_crc32 23%, UUValidData 11%
4
5	- include ecb.h to deal with compiler builtins and endianness.
6	- some further µ-optimisations in hot code, especially for yEnc.
7	- replace crc32 function by slice-by-16 version by Stephan Brumme,
8	which should speed up yEnc en-/decoding.
9	- yEnc: do not calculate two crcs per part, instead, combine
10	the part crcs together to form the file crc.
11	- yEnc: allow pcrc= in addition to pcrc32= for yenc trailers, as
12	some draft mentions both and it is actually in active use.
13	- yEnc: ignore crc32= on multiparts, except on the last part.
14	- be more precise in documenting code licenses in COPYING.
15
16	1.62 Mon Feb 17 23:19:42 CET 2020
17	- major performance improvement by simplifying code in _FP_gets
18	to not use fscanf. This might slow things down on platforms
19	with very slow fgetc.
20
21	1.61 Sun Feb 9 18:38:29 CET 2020
22	- lint uulib: fix some format string type mismatches
23	and some other minor issues.
24
25	1.6 Thu Oct 24 17:11:54 CEST 2019
26	- fix heap overflow (testcase by Noel Duffy, reported
27	by Robert Scheck). The defense-in-depth mechanism based
28	on mmap should make this unexploitable for other than denial
29	of service, on systems supporting mmap/mprotect.
30
31	1.5 Sat Jul 11 03:56:06 CEST 2015
32	- fix a heap overflow (testcase by Krzysztof Wojtaś).
33	- on systems that support it (posix + mmap + map_anonymous),
34	allocate all dynamic areas via mmap and put four guard
35	pages around them, to catch similar heap overflows
36	safely in the future.
37	- find a safer way to pass in CC/CFLAGS to uulib.
38	- added stability canary support.
39
40	1.4 Sun May 29 17:17:01 CEST 2011
41	- avoid a classical buffer overflow in case a progress
42	message is too long.
43	- this release adds dependencies for snprintf/vsnprintf.
44	- some uuencode encoders do not generate a final "space" line
45	before the "end" marker, so do not rely on the line to be there.
46
47	1.34 Tue Dec 14 22:20:00 CET 2010
48	- fix a one-byte-past-end-write buffer overflow in UURepairData
49	(reported, analysed and testcase provided by Marco Walther).
50	- quoted-printable decoding was completely broken, try a fix.
51
52	1.33 Wed Oct 28 09:04:38 CET 2009
53	- handle yEnc files with part end=0 and total= more gracefully.
54	I wish yEnc had been created by somebody who knows;
55	what he does;
56	but I doubt he even knows;
57	what he did.
58
59	1.32 Wed Sep 16 20:07:13 CEST 2009
60	- Due to a glitch with CVS, configure lacked executable bits.
61	(Quickly reported by Anton Berezin).
62
63	1.31 Wed Sep 16 09:04:30 CEST 2009
64	- do not use system-replacements for case-insensitive string
65	functions when found, as they are broken on too many systems
66	(mostly bsds, as usual, but at least some versions of GNU/Linux
67	disagree with themselves apparently). Analyzed by Anton Berezin.
68
69	1.3 Sat Aug 29 01:24:35 CEST 2009
70	- major changes, new bugs and changes in decoding behaviour are
71	expected (but not intended).
72	- major scanning and decoding speed-up (by a factor of 4),
73	by replacing ultra-slow _FP_gets and improving IsKnownHeader
74	(but fgets is still responsible for >50% if the time).
75	- new option OPT_AUTOCHECK to disable O(n) UUCheckGlobalList
76	call after every loadfile, majorly speeds up large decodes
77	(easily by a factor of 10..100).
78	- allow "Smerge -1" to call UUCheckGlobalList.
79	- majorly speed up part insertion (still O(n), but much faster).
80	- allow for 1023 octet headers instead of the standard
81	255 octet ones.
82	- support strcasestr, strcasecmp, strncasecmp for added speed.
83
84	1.12 Mon Oct 13 14:11:01 CEST 2008
85	- use the yencode filesize as additional matching criterium
86	to avoid false matches.
87	- made the example decoder more verbose w.r.t. error handling.
88	- removed potentially confusing decode_temp calls from
89	example decoder.
90
91	1.11 Fri Jun 13 15:32:30 CEST 2008
92	- don't ask.
93
94	1.10 Fri Jun 13 14:22:42 CEST 2008
95	- fix an infinite-looping problem when scanning in freestyle
96	mode (testcase provided by Pieter Geens and Reinhard Pfau).
97
98	1.09 Fri May 25 19:38:11 CEST 2007
99	- create something sensible, trust a windows program to fuck
100	it up: work around literal "(null)" filenames in yenc-encoded
101	files.
102	- some minor cleanups.
103
104	1.08 Sat Dec 16 23:27:13 CET 2006
105	- URGENT update, the last release did not
106	decode files correctly, usually not at all.
107	- my last patch was, of course, completely bogus.
108	(sorry. looked simple...).
109
110	1.07 Sun Dec 10 17:41:46 CET 2006
111	- fixed an uninitialised variable based on analysis
112	and patch by Mark Martinec.
113
114	1.06 Tue Dec 6 00:56:05 CET 2005
115	- fix a number of int/long format errors in the encoding part and
116	fix some signed/unsigned char problems of unknown relevance,
117	reported by Jonas Smedegaard.
118	- new EXPERIMENTAL options OPT_RBUF and OPT_WBUF to set default
119	stdio buffer size for reading and writing files.
120
121	1.051 Thu Mar 3 18:00:52 CET 2005
122	- change of contact address.
123	- updated perl parts to GPLv2.
124
125	1.05 Fri Feb 25 22:50:27 CET 2005
126	- fix a (likely exploitable) segfault problem, (tracked down
127	and/or reported by Mark Martinec and Robert Lewis).
128
129	1.04 Tue Dec 28 15:08:44 CET 2004
130	- slightly improved subject filename extraction.
131	Also see the filename callback in the example-decoder.
132
133	1.03 Sun Apr 18 22:05:43 CEST 2004
134	- upgrade to uudeview-0.5.20, which supposedly fixes a few buffer
135	overflows. However, judging from the patch these vulnerabilities
136	were not present in the Convert-UUlib version (I might err, though,
137	documentation on the actual exploits is scarce).
138	As every uudeview update usually brings more instability than stability
139	I advise against using this version until it has proven itself
140	to be stable, or 1.02 is proven to contain the same buffer overflows.
141
142	1.02 Sun Apr 18 16:47:26 CEST 2004
143	- renamed crc32 to uulib_crc32, to work around yet another
144	shortcoming in this oh-so-outdated macosx.
145
146	1.01 Sun Feb 1 19:49:51 CET 2004
147	- make it compile with 5.005_03 (reported by Anton Berezin).
148	- very short files might not be detected because the header
149	parsing code might skip them.
150
151	1.0 Thu Nov 6 14:32:08 CET 2003
152	- change bracket policy to give priority to (x/y) over (x),
153	and use the last bracket found otherwise.
154	- part numbers at the end of the subject are now recognized.
155	- updated to uulib-0.5.19. Differences and bugfixes kept.
156
157	0.31 Wed Oct 16 01:22:34 CEST 2002
158	- no internal code changes.
159	- much better documentation, now considered "useful".
160	- removed the procedural interface.
161	- fought the wish to perl-5.8'ify and thus simplify the code :().
162	- stress-tested version 0.3 against 70000 data postings since sunday.
163
164	0.3 Sun Oct 13 15:12:11 CEST 2002
165	- updated to uulib 0.5.18. As expected, some but not all of my bugfixes
166	have went into uulib, so the number of differences decreased a bit
167	again.
168	- I found out that the library is being sold commercially by Frank
169	Pilhofer, disregarding the GPL and ignoring the rights of the people
170	who sent in patches :(
171	- vastly more useful documentation in the .pm file.
172	- much smaller distribution filesize ;)
173
174	0.213 Sat Jul 27 21:16:30 CEST 2002
175	- fixed another buffer overflow, also added a santity check to fgets.
176
177	0.212 Sat Apr 6 03:52:13 CEST 2002
178	- fix yEnc decoding support.
179	- new option OPT_DOTDOT, that makes uulib unescape dot-escaping.
180	- increased linelength to 1195.
181
182	0.211 Fri Apr 5 23:56:46 CEST 2002
183	- sigh. new version of uulib, new braindamaged buffer overflows
184	fixed. Diffs like "line length 256 => 300" should have
185	alerted me...
186
187	0.21 Sun Mar 31 22:06:05 CEST 2002
188	- bumped uulib from 0.5.15 to 0.5.17 (adds some yEnc support, YENC_ENCODED).
189	- source-renamed FP_-symbols to _FP_ again, but still define them
190	to be FP_* in the object file.
191	- new callback: SetFileNameCallback (actually documented ;)
192
193	0.201 Sun Sep 16 03:43:38 CEST 2001
194	- fixed another segfault-bug in uulib, thanks to Lars Hecking who
195	provides all the test files ;)
196
197	0.2 Thu Jun 14 18:42:49 CEST 2001
198	- Frank Pilhofer is NOT dead! He is alive, kicking, and fixing
199	bugs in uulib now! Since his uulib and mine diverged quite
200	a bit I am still in the process of merging his bugfixes
201	into my version. So be aware that this might introduce new
202	instabilities...
203	- merged uulib-0.5.15
204	- optimize decode_file to not copy the file contents when a rename
205	succeeds (TMPDIR same disk as destination).
206	- updated documentation.
207
208	0.111 Fri May 4 22:29:09 CEST 2001
209	- heuristic to decode microsoft-rfc822-headers (microsoft uses
210	'\' as path seperator which is also the rfc822 quote character.
211	of course they could have quoted it, but it's microsoft, so
212	only the devil knows how they try to parse mime (they obviously
213	haven't read the standard AGAIN). in case you wonder, this fixes
214	the "c:my_documentssend.doc" "filename" problem.
215
216	0.11 Sun Jul 16 22:42:57 CEST 2000
217	- fixed a bug in the file callback.
218
219	0.10 Tue May 2 03:31:09 CEST 2000
220	- fixed one very longstanding bug that might have been the cause
221	for the many mysterious segmentation faults. Hopefully.
222	- improved documentation to the point that it slowly becomes useful
223	as a (very quick) reference.
224
225	0.07 Wed Nov 24 18:45:23 CET 1999
226	- removed END { CleanUp } from UUlib.pm, perl is too
227	buggy to rely on this in the face of popen etc..
228
229	0.06 Thu Jul 29 22:48:18 CEST 1999
230	- Thierry Bezecourt had the very reasonable idea to include
231	the example source in the manpage.
232
233	0.05 Mon May 24 11:37:25 CEST 1999
234	- LoadFile returns number of parts loaded.
235
236	0.04 Mon May 10 22:02:31 CEST 1999
237	- fixed a couple of portability problems.
238	- UUEncodeToStream had a wrong definition.
239	- added primitive testsuite.
240	- added README.
241
242	0.03 Sat May 8 22:40:01 CEST 1999
243	- renamed it to Convert::UUlib.
244
245	0.02 Fri May 7 01:26:19 CEST 1999
246	- function names exportable (use :all-tag).
247	- made uulist-related functions methods.
248	- the first fix from cpan-tester. sigh you people save the
249	world, continuously.
250
251	0.01 Thu May 6 14:30:44 1999
252	- original version; created by h2xs 1.19
253