Created on 2008-03-29 13:38 by nedbat, last changed 2015-10-22 18:37 by brett.cannon. This issue is now closed.
|continue.py||nedbat, 2008-03-29 13:38||Test file to trace to see problem.|
|msg64692 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2008-03-29 13:38|
When tracing line execution with sys.settrace, a particular code structure fails to report an executed line. The line is a continue statement after an if condition in which the if condition is true every time it is executed. Attached is a file with two copies of the same code, except in the first the if condition is always true, and in the second it is sometimes true. In the first, trace.py reports that the continue is never executed, even though it is (as evidenced from the values of a, b, and c after execution). In the second code, the continue is properly reported. This bug has been present since version 2.3. 2.2 does not exhibit it (trace.py didn't exist in 2.2, but coverage.py shows the problem also). To see the problem, execute "trace.py -c -m continue.py". Then continue.py.cover will show: 1: a = b = c = 0 101: for n in range(100): 100: if n % 2: 50: if n % 4: 50: a += 1 >>>>>> continue else: 50: b += 1 50: c += 1 1: assert a == 50 and b == 50 and c == 50 1: a = b = c = 0 101: for n in range(100): 100: if n % 2: 50: if n % 3: 33: a += 1 17: continue else: 50: b += 1 50: c += 1 1: assert a == 33 and b == 50 and c == 50
|msg64699 - (view)||Author: Amaury Forgeot d'Arc (amaury.forgeotdarc) *||Date: 2008-03-29 14:41|
This is because of a "peephole" optimization of the generated bytecode: a jump instruction which target is another jump instruction can be modified modified to target the final location. You gain a few opcodes, but tracing is confusing... Not sure how to fix this, though.
|msg64713 - (view)||Author: Alexander Belopolsky (belopolsky) *||Date: 2008-03-29 18:25|
I think this is not a bug. Here is a simpler way to illustrate the issue: def f(x): for i in range(10): if x: pass continue f(True) f(False) If you run the code above under trace, you get the following coverage: 1: def f(x): 22: for i in range(10): 20: if x: 10: pass 10: continue 1: f(True) 1: f(False) Note that the 'continue' line is executed 10 instead of expected 20 times. This happens exactly as Amaury explained. If you disassemble f, you'll see 2 0 SETUP_LOOP 34 (to 37) 3 LOAD_GLOBAL 0 (range) 6 LOAD_CONST 1 (10) 9 CALL_FUNCTION 1 12 GET_ITER >> 13 FOR_ITER 20 (to 36) 16 STORE_FAST 1 (i) 3 19 LOAD_FAST 0 (x) 22 JUMP_IF_FALSE 4 (to 29) 25 POP_TOP 4 26 JUMP_ABSOLUTE 13 >> 29 POP_TOP 5 30 JUMP_ABSOLUTE 13 33 JUMP_ABSOLUTE 13 >> 36 POP_BLOCK >> 37 LOAD_CONST 0 (None) 40 RETURN_VALUE Note how peephole optimizer replaced jump to the 'continue' line (5) from the 'pass' line (4) with a jump to the 'for' line by replacing 4 26 JUMP_FORWARD 1 (to 30) with 4 26 JUMP_ABSOLUTE 13 I say this is not a bug because trace is correct in showing that the continue statement is never reached when executing f(True).
|msg64716 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2008-03-29 18:51|
I see that the cause of the problem is the peephole optimizer. That doesn't mean this isn't a problem. I am measuring the code coverage of a set of tests, and one of my lines is being marked as not executed. This is not the fault of the tests, because in fact, without the optimization, the line would be executed. Conceptually, the line has been executed (the loop is restarted, rather than execution continuing). I don't know what the solution to this is. Some options include fixing the line tracing code to somehow indicate that the continue was executed; or providing a way to disable peephole optimization for times when accurate execution tracing is more important than speed.
|msg64720 - (view)||Author: Alexander Belopolsky (belopolsky) *||Date: 2008-03-29 19:42|
On Sat, Mar 29, 2008 at 2:51 PM, Ned Batchelder <email@example.com> wrote: > > Ned Batchelder <firstname.lastname@example.org> added the comment: > > I am measuring the code coverage of a set of tests, and one of my lines > is being marked as not executed. This is not the fault of the tests, > because in fact, without the optimization, the line would be executed. > Conceptually, the line has been executed (the loop is restarted, rather > than execution continuing). > .. but the continue statement on line 5 is NOT executed in x == True case. Note that without optimization, the if statement + the continue line translate to 3 19 LOAD_FAST 0 (x) 22 JUMP_IF_FALSE 4 (to 29) 25 POP_TOP 4 26 JUMP_FORWARD 1 (to 30) >> 29 POP_TOP 5 >> 30 JUMP_ABSOLUTE 13 where the second jump is to the continue statement. Peephole optimizer recognizes that the jump target is an unconditional jump and changes the code to jump directly to the final target bypassing the continue line. The optimized code is 3 19 LOAD_FAST 0 (x) 22 JUMP_IF_FALSE 4 (to 29) 25 POP_TOP 4 26 JUMP_ABSOLUTE 13 >> 29 POP_TOP 5 30 JUMP_ABSOLUTE 13 If x is true, line five is NOT executed. > I don't know what the solution to this is. Some options include fixing > the line tracing code to somehow indicate that the continue was > executed; or providing a way to disable peephole optimization for times > when accurate execution tracing is more important than speed. > I think it is a good idea to provide a way to disable peephole optimizer. In fact, I recently proposed exactly that in msg64638. My only problem is that I would like to follow gcc tradition and make -O option take an optional numeric argument with 0 meaning no optimization and increasingly aggressive optimization as the argument increases. Unfortunately -O0 will be confusingly similar to -OO. Since -OO is not really optimization, but rather "strip" option, it should probably migrate to -s or something. In any case, such drastic changes to command line options are not acceptable for 2.x, but maybe possible for 3.0. I can easily implement -N (No optimization) or -g (debug) option that will disable the peephole optimizer if there is support for such feature.
|msg64725 - (view)||Author: Amaury Forgeot d'Arc (amaury.forgeotdarc) *||Date: 2008-03-29 20:50|
> Unfortunately -O0 will be confusingly similar to -OO. On my browser, both are shown identically at the pixel level. Microsoft compilers use -Od to disable optimization...
|msg64726 - (view)||Author: Raymond Hettinger (rhettinger) *||Date: 2008-03-29 20:58|
This has basically almost never been a problem in the real world. No need to complicate the world further by adding yet another option and the accompanying implementation-specific knowledge of why you would ever want to use it. Also, when the peepholer is moved (after the AST is created, but before the opcodes), then little oddities like this will go away. Recommend closing as "won't fix".
|msg64729 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2008-03-29 21:19|
I recognize that this is an unusual case, but it did come up in the real world. I found this while measuring test coverage, and the continue line was marked as not executed, when it was. I don't understand "when the peepholer is moved", so maybe you are right that this will no longer be an issue. But it seems to me to be endemic to code optimization to lose the one-to-one correspondence between source lines and ranges of bytecodes. And as the compiler becomes more complex and performs more optmizations, problems like this will likely increase, no? In any case, I'd like to know more about the changes planned for the AST and compiler...
|msg64754 - (view)||Author: Alexander Belopolsky (belopolsky) *||Date: 2008-03-30 17:39|
On Sat, Mar 29, 2008 at 4:58 PM, Raymond Hettinger <email@example.com> wrote: > This has basically almost never been a problem in the real world. I believe Ned gave an important use case. In coverage testing, optimized runs can show false gaps in coverage. In addition, a no optimize option would provide a valuable learning tool. Python has an excellent simple VM very suitable for a case study in introductory CS courses. Unfortunately, inability to disable peephole optimizer makes understanding the resulting bytecode more difficult, particularly given some arbitrary choices made by the optimizer (such as 2*3+1 => 7, but 1+2*3 => 1+6). Furthermore, as Raymond suggested in another thread, peephole optimizer was deliberately kept to bare minimum out of concerns about compilation time. Given that most python code is pre-compiled, I think it is a rare case when code size/speed improvements would not be worth increased compilation time. In a rare case when compilation time is an issue, users can consider disabling optimization. Finally, an easy way to disable the optimizer would help in developing the optimizer itself by providing an easy way to measure improvements and debugging. > No need to complicate the world further by adding yet another option and > the accompanying implementation-specific knowledge of why you would > ever want to use it. > This would not really be a new option. Most users expect varying levels of optimization with -O option and python already has 3 levels: plain, -O, and -OO or Py_OptimizeFlag = 0,1, and 2. Moreover, in fact, Py_OptimizeFlag can be set to an arbitrary positive integer using undocumented -OOO.. option. I don't see how anyone would consider adding say -G with Py_OptimizeFlag = -1 that would disable all optimization as "complicating the world." > Also, when the peepholer is moved (after the AST is created, but before > the opcodes), then little oddities like this will go away. > I don't see how moving optimization up the chain will help with this particular issue. Note that the problem is not with peepholer emiting erroneous line number information, but the fact that the continue statement is optimized away by replacing the if statement's jump to continue with a direct jump to the start of the loop. As I stated in my first comment, trace output is correct and as long as the compiler avoids redundant double jumps, the continue statement will not show up in trace regardless where in compilation chain it is optimized. The only way to get correct coverage information is to disable double jump optimization.
|msg64764 - (view)||Author: Raymond Hettinger (rhettinger) *||Date: 2008-03-30 21:01|
Weigh the cost/benefit carefully before pushing further. I don't doubt the legitimacy of the use case, but do think it affects far fewer than one percent of Python programmers. In contrast, introducing new command line options is a big deal and will cause its own issues (possibly needing its own buildbot runs to exercise the non-optimized version, having optimized code possibly have subtle differences from the code being traced/debugged/profiled, and more importantly the mental overhead of having to learn what it is, why it's there, and when to use it). My feeling is that adding a new compiler option using a cannon to kill a mosquito. If you decide to press the case for this one, it should go to python-dev since command line options affect everyone. This little buglet has been around since Py2.3. That we're only hearing about it now is a pretty good indicator that this is a very minor in the Python world and doesn't warrant a heavy-weight solution. It would be *much* more useful to direct effort improving the mis- reporting of the number of arguments given versus those required for instance methods: >>> a.f(1, 2) TypeError: f() takes exactly 1 argument (3 given)
|msg64765 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2008-03-30 21:37|
Raymond, do you have a cannon-less recommendation of how to kill this particular mosquito?
|msg64768 - (view)||Author: Alexander Belopolsky (belopolsky) *||Date: 2008-03-30 22:24|
On Sun, Mar 30, 2008 at 5:01 PM, Raymond Hettinger <firstname.lastname@example.org> wrote: .. > It would be *much* more useful to direct effort improving the mis- > reporting of the number of arguments given versus those required for > instance methods: > >>> a.f(1, 2) > TypeError: f() takes exactly 1 argument (3 given) Please see issue2516.
|msg64774 - (view)||Author: Alexander Belopolsky (belopolsky) *||Date: 2008-03-31 01:34|
On Sun, Mar 30, 2008 at 5:01 PM, Raymond Hettinger <email@example.com> wrote: .. > Weigh the cost/benefit carefully before pushing further. I don't doubt > the legitimacy of the use case, but do think it affects far fewer than > one percent of Python programmers. I agree with you, but only because fewer than 1% of Python programmers have complete test coverage for their code. :-) On the other hand, I wanted a no-optimize option regardless of the trace issue. Once it is there, I am sure everyone interested in how python compiler works will use it. (I am not sure what % of Python programmers would fall into that category.) I don't know how big of a deal an extra buildbot is, but I don't think it will be necessary. It is hard to imagine optimization that would fix (mask) errors in non-optimized code. Therefore, a non-optimized buildbot is unlikely to flag errors that ar not present in optimized runs. On the other hand errors introduced by optimizer will be easier to diagnose if they disappear when the code runs without optimization. Mental overhead is important, but I think it will be easier to explain the effect of no optimize option than to explain what -O does in the current version. As far as I can tell, -O has nothing to do with peephole optimization and only removes assert statements and replaces __debug__ with 0. I am sure most python users are not aware of the fact that peephole optimization is performed without -O option. > My feeling is that adding a new compiler option using a cannon to kill > a mosquito. If you decide to press the case for this one, it should go > to python-dev since command line options affect everyone. > As an alternative to the command line option, what would you say to making sys.flags.optimize writeable and disable peepholer if Py_OprimizeFlag < 0? This will allow python tracing tools to disable optimization from within python code. The fact that setting sys.flags.optimize flag will not affect modules that are already loaded is probably a good thing because tracing code itself will run optimized. Such tracing tools may also need to use a custom importer that would ignore precompiled code and effectively set dont_write_bytecode flag. > This little buglet has been around since Py2.3. That we're only > hearing about it now is a pretty good indicator that this is a very > minor in the Python world and doesn't warrant a heavy-weight solution. > I still maintain that this is not a bug. Not hearing about it before is probably an indication that users sophisticated enough to try to achieve full test coverage for their code were able to recognize false coverage gaps as such.
|msg64775 - (view)||Author: Raymond Hettinger (rhettinger) *||Date: 2008-03-31 01:59|
Marking this one as closed. Also, rejecting the various ways to disable peephole optimization. This was discussed with Guido long ago and the decision essentially recognized that for most practical purposes the output of the peepholer is the generated code and no good would come from exposing upstream intermediate steps. Since then, I believe Neal got Guido's approval for either the -O or - OO option to generate new optimizations that potentially change semantics. In that situation, there is a worthwhile reason for the enable/disable option.
|msg64809 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2008-04-01 13:23|
It's hard for me to agree with your assessment that for no practical good would come from disabling the optimizer. Broadly speaking, there are two types of code execution: the vast majority of the time, you execute the code so that it can do its work. In this case, speed is most important, and the peephole optimizer is a good thing. But another important case is when you need to reason about the code. This second case includes coverage testing, debugging, and other types of analysis. Compiled languages have long recognized the need for both types of compilation, which is why they support disabling optimization entirely. As Python becomes more complex, and more broadly deployed, the needs of the two types of execution will diverge more and more. More complex optimizations will be attempted in order to squeeze out every last drop of performance. And more complex tools to reason about the code will be developed to provide rich support to those using Python for complex development. I see discussion here of moving the optimizer to the AST level instead of the bytecode level. This won't change the situation. The optimizer will still interfere with analysis tools. As a developer of analysis tools, what should I tell my users when their code behaves mysteriously?
|msg64957 - (view)||Author: Terry J. Reedy (terry.reedy) *||Date: 2008-04-05 02:05|
While I agree with Raymond that the interpreter should be left alone, this could be reclassified (and reopened) as a doc issue. The current trace doc (Lib Ref 25.10) says rather tersely "The trace module allows you to trace program execution, generate annotated statement coverage listings, print caller/callee relationships and list functions executed during a program run." This could be augmented with a general statement that the effect of certain statements may get computed during compilation and not appear in the runtime trace -- or a more specific statement about continue, break, and whatever else. AS for continue.py, it seems that the apparent non-execution of a continue line indicates one of two possible problems. 1. The if statement is equivalent to 'if True:', at least for the intended domain of input, hence redundant, and hence could/should be removed. 2. Otherwise, the inputs are incomplete as far as testing the effect of not taking the if-branch, and hence could/should be augmented. Either way, it seems to me that the lack of runtime execution of continue, coupled with better documentation, could usefully point to possible action.
|msg140281 - (view)||Author: Jean-Paul Calderone (exarkun) *||Date: 2011-07-13 16:07|
Since the main argument for not fixing this bug seems to be that it doesn't affect many users, it seems like I should comment here that the issue is affecting me. A recently proposed addition to Twisted gets bitten by this case, resulting in a report of less than full test coverage when in fact the tests do exercise every line and branch of the change. Perhaps it is too hard to add and maintain a no-optimizations feature for Python (although I agree with Ned that this would be a useful feature for many reasons, not just to fix this bug). There are other possible solutions to the issue of inaccurate coverage reports though. For example, Python could provide an API for determining which lines have code that might be executed. coverage.py (and the stdlib trace.py) currently use the code object's lnotab to decide which lines might be executable. Maybe that should omit "continue" lines that get jumped over. If the line will never execute, it seems there is no need to have it in the lnotab. Using the lnotab is something of a hack though, so it might also make sense to leave it alone but introduce an API to get the same information, but corrected for whatever peephole optimizations the interpreter happens to have. As far as the "not a bug" arguments go, I don't think it matters much whether you ultimately decide to call it a bug or a feature request. It *is* clearly a useful feature to some people though, and rejecting the requested behavior as "not a bug" doesn't help anyone. So call it a feature request if that makes it more palletable. :)
|msg140290 - (view)||Author: Terry J. Reedy (terry.reedy) *||Date: 2011-07-13 17:04|
I think supporters of this feature request should take discussion to python-ideas to try to gather more support. The initial post should summarize reasons for the request, possible implementations, and the counter-arguments of Raymond.
|msg140303 - (view)||Author: Terry J. Reedy (terry.reedy) *||Date: 2011-07-13 20:29|
Choose pydev if you want. Discussion there is *usually* (but definitely not always) more focused on implementation of uncontroversial changes. I am pretty much +-0 on the issue, though Jean-Paul's post seems to add to the + side arguments that might be persuasive to others also.
|msg218811 - (view)||Author: Raymond Hettinger (rhettinger) *||Date: 2014-05-19 19:55|
There has been no activity on this for several year. Marking as rejected for the reasons originally listed.
|msg218814 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2014-05-19 20:26|
Raymond, thanks for keeping us honest! I am still hoping to convince people that this is a good idea. I think Guido's +1 (https://mail.python.org/pipermail/python-dev/2012-December/123099.html) should help in that regard. Part of your reason for today's rejection is the lack of activity. Can I assume that with a patch you would be supportive?
|msg218816 - (view)||Author: Trip Volpe (Trip.Volpe)||Date: 2014-05-19 20:34|
I found this issue just the other day while researching why we were getting false gaps in our test coverage reports (using Ned's coverage module, natch!). I agree that this seems like a fairly minor nuisance, but it's a nuisance that anybody who has tests and measures test coverage will run into sooner or later -- and that's *everybody*, right? I think some kind of fix ought to be discussed. After all, "it should be possible to have accurate coverage results" is a proposition that seems fairly reasonable to me.
|msg218829 - (view)||Author: Terry J. Reedy (terry.reedy) *||Date: 2014-05-20 02:17|
Trip, see msg140290, which was ignored.
|msg218869 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2014-05-21 11:28|
Python-Ideas thread started: https://mail.python.org/pipermail/python-ideas/2014-May/027893.html
|msg218891 - (view)||Author: Raymond Hettinger (rhettinger) *||Date: 2014-05-22 05:49|
Ned, why is your proposal to turn-off ALL peephole transformations with COMMAND-LINE switch? * Why not just turn-off the jump-to-jump? Do you really need to disable constant folding and other transformations? * Have you explored whether the peephole.c code can be changed to indicate the continue-statement was visited? * Why does this have to be a command-line setting rather than a flag or environment variable settable by coverage.py? * Is there some less radical way the coverage.py can be taught to make the continue-statement as visited? * Are you requesting that optimization constraints be placed on all of the implementations of Python (Jython, PyPy, and IronPython) to make coverage.py perfect? * Do you want to place limits on what can be done by Victor's proposed AST tranformations which will occur upstream from the peepholer and will make higher level semantically-neutral transformations *prior* to code generation. * Have you considered whether the genererated PYC files need a different magic number or some other way to indicate that they aren't production code? * If coverage.py produces a report on different code than the production run, doesn't that undermine some of the confidence the meaningfulness of the report? In other words, are you sure that you're making the right request and that it is really worth it? Do we really have to open this can of worms to make coverage.py happy?
|msg218892 - (view)||Author: STINNER Victor (haypo) *||Date: 2014-05-22 07:12|
> Have you considered whether the genererated PYC files need a different magic number or some other way to indicate that they aren't production code? Would it make sense to use a different sys.implementation.cache_tag? For example, the tag si currently "cpython-35". We can use "cpython-35P" when peephole optimizations are disabled. So you can have separated .pyc and .pyo files and the disabling peephole optimizations is compatible with -O and -OO command line options.
|msg218893 - (view)||Author: STINNER Victor (haypo) *||Date: 2014-05-22 10:06|
Oh, another option to solve the .pyc file issue is to *not* write .pyc files if the peephole optimizer is disabled. If you disable an optimizer, you probably don't care of performances.
|msg218894 - (view)||Author: Ned Batchelder (nedbat) *||Date: 2014-05-22 10:46|
I thought we were discussing this on Python-Ideas?
|msg218923 - (view)||Author: Raymond Hettinger (rhettinger) *||Date: 2014-05-22 22:21|
[Victor] > Oh, another option to solve the .pyc file issue is to *not* > write .pyc files if the peephole optimizer is disabled. > If you disable an optimizer, you probably don't care of performances. That is an inspired idea and would help address one of the possible problems that could be caused by a new on/off switch.
|msg221250 - (view)||Author: Pedro Gimeno (pgimeno)||Date: 2014-06-22 13:10|
I consider peephole optimization when no optimization was requested a bug. Documentation for -O says it "Turns on basic optimizations". Peephole optimization is a basic optimization, yet it is performed even when no basic optimizations were requested. No need to add a switch. Just don't optimize if not requested.
|msg253345 - (view)||Author: Brett Cannon (brett.cannon) *||Date: 2015-10-22 18:37|
I believe the python-ideas thread on this topic came to the conclusion that a -X flag -- e.g., `-X DisableOptimizations` -- would be a good way to turn off all optimizations. The flag could then either blindly set sys.dont_write_bytecode to True or set sys.flags.optimize to -1 in which case a bytecode file named e.g. foo.cpython-36.opt--1.pyc would be written which won't lead to any conflicts (I wish we could use False for sys.flags.optimize but that has the same values as 0 which is the default optimization level). Does one of those proposal seems acceptable to everyone? Do people like Ned who asked for this feature have a preference as to whether the bytecode is or is not written out to a .pyc file?
messages: + msg253345
messages: + msg221250
|2014-05-22 22:21:59||rhettinger||set||messages: + msg218923|
|2014-05-22 10:46:25||nedbat||set||messages: + msg218894|
|2014-05-22 10:06:05||haypo||set||messages: + msg218893|
messages: + msg218892
versions: + Python 3.5, - Python 3.4
|2014-05-21 11:28:29||nedbat||set||messages: + msg218869|
|2014-05-20 02:17:01||terry.reedy||set||messages: + msg218829|
|2014-05-19 20:34:59||Trip.Volpe||set||messages: + msg218816|
|2014-05-19 20:26:31||nedbat||set||messages: + msg218814|
|2014-05-19 19:55:14||rhettinger||set||status: open -> closed|
messages: + msg218811
versions: + Python 3.4, - Python 3.3
|2011-07-15 14:43:16||eric.araujo||set||title: Add mechanism to diasable optimizations -> Add mechanism to disable optimizations|
|2011-07-13 20:32:41||terry.reedy||set||messages: - msg140304|
|2011-07-13 20:29:40||terry.reedy||set||messages: + msg140304|
|2011-07-13 20:29:38||terry.reedy||set||messages: + msg140303|
|2011-07-13 17:04:00||terry.reedy||set||type: behavior -> enhancement|
title: Line tracing of continue after always-taken if is incorrect -> Add mechanism to diasable optimizations
messages: + msg140290
versions: + Python 3.3, - Python 2.6, Python 2.5, Python 2.4, Python 2.3
|2011-07-13 16:07:51||exarkun||set||status: closed -> open|
nosy: + exarkun
messages: + msg140281
resolution: wont fix -> (no value)
messages: + msg64957
|2008-04-01 13:23:15||nedbat||set||messages: + msg64809|
|2008-03-31 01:59:38||rhettinger||set||status: open -> closed|
resolution: wont fix
messages: + msg64775
|2008-03-31 01:34:08||belopolsky||set||messages: + msg64774|
|2008-03-30 22:24:43||belopolsky||set||messages: + msg64768|
|2008-03-30 21:37:34||nedbat||set||messages: + msg64765|
|2008-03-30 21:01:13||rhettinger||set||messages: + msg64764|
|2008-03-30 17:39:10||belopolsky||set||messages: + msg64754|
|2008-03-30 04:00:30||ajaksu2||set||nosy: + ajaksu2|
|2008-03-29 21:19:28||nedbat||set||messages: + msg64729|
messages: + msg64726
|2008-03-29 20:50:40||amaury.forgeotdarc||set||messages: + msg64725|
|2008-03-29 19:42:19||belopolsky||set||messages: + msg64720|
|2008-03-29 18:51:54||nedbat||set||messages: + msg64716|
messages: + msg64713
messages: + msg64699