00:03:46 Ahah! 00:03:48 Gato works. 00:03:56 Maybe I can write the Andrei Machine 9000! 00:04:10 Slereah: zgoshmgz 00:05:49 I am full of hope. 00:06:01 I really don't want to write it in Python. 00:06:13 Slereah: whuts gato 00:06:46 http://gato.sourceforge.net/http://gato.sourceforge.net/? 00:06:49 It's a program to write algorithms on graphs. 00:06:55 Yes, that one. 00:06:59 Slereah: im sorry but that code is python 00:07:01 they just embed it 00:07:05 and add their own functions 00:07:25 Yes. 00:07:37 But what I mean is, I don't want to build everything 00:07:57 I am terrible at coding, so I really don't want to write the whole graph gestion myself 00:08:27 Plus, all my interpreters have some way of seeing what happens to the machine, and I don't want to make an exception. 00:09:19 Slereah: Cool. 00:14:17 -!- timotiis has quit ("rebloodybooting! They said I wouldn't have to do this on linux!"). 00:14:27 hah 00:16:41 -!- timotiis has joined. 00:18:20 timotiis: sorry 00:18:45 hehe 00:39:46 http://rafb.net/p/dUgDuS45.html 00:39:53 looks like ccbi is broken Deewiant ^ 00:50:03 -!- Sgeo has joined. 00:53:14 -!- timotiis has quit (Connection timed out). 01:01:58 -!- Sgeo has quit (Read error: 104 (Connection reset by peer)). 01:05:20 -!- Sgeo has joined. 01:17:05 On Fark: 01:17:06 "There are actually people named null, and yes, I've actually seen this cause problems in production database code. 'Null' was the last name where I saw it however." 01:26:31 hmm 01:27:10 public class Closure { public Closure(Object... vars); public U apply(T arg); } 01:27:15 e.g. 01:27:45 new Closure(a, b, c) { public Integer apply(Integer n) { return n + (Integer)this.vars[0]; } } 01:27:51 this.vars={a,b,c} 01:28:01 maybe nice if you are compiling a language to jvm 01:55:14 hah! ehird: i was not there! 01:55:25 *here 01:55:29 oklofok: heh 01:57:13 oklofok: but ... yeah 01:57:22 that's also why you shouldnt compile oklotalk-- to the jvm, btw 01:57:25 it doesn't like closures 02:00:10 i would probably make the compilation at a bit lower level 02:00:29 but python bytecode might be nicer 02:00:47 also, as i don't really like java, i don't really wanna use it's bytecode 02:00:48 :o 02:00:55 anyway, /me goes back 02:01:04 before you start asking again :-) 02:01:06 -> 02:27:05 -!- Quendus has quit (Read error: 113 (No route to host)). 02:27:42 -!- Quendus has joined. 02:30:10 quendus 02:30:12 haven't seen that name 02:30:13 hi 04:06:47 -!- ehird has quit (Read error: 104 (Connection reset by peer)). 04:25:02 The Security Manager is vetoing my using Firebug in Fx3.. 04:25:26 >>> 2+2 04:25:26 commandLine.evaluate FAILS: [Exception... "Security Manager vetoed action" nsresult: "0x80570027 (NS_ERROR_XPC_SECURITY_MANAGER_VETO)" location: "JS frame :: chrome://firebug/content/commandLine.js :: :: line 100" data: no] 04:26:35 -!- fizzie has quit (Remote closed the connection). 04:26:39 -!- fizzie has joined. 04:38:56 -!- adu has joined. 05:00:34 lol at http://youtube.com/watch?v=g3KufmUlF5M 05:00:39 (I'm commented SJGster) 05:38:45 -!- calamari has joined. 05:38:50 hi 05:39:36 Hi calamari 05:39:41 http://youtube.com/watch?v=g3KufmUlF5M == fail 05:44:37 hi 06:20:58 -!- calamari has quit ("Leaving"). 06:38:41 AnMaster: cheers, I'll take a look tonight 06:45:44 -!- olsner has joined. 06:47:42 -!- Sgeo has quit ("Ex-Chat"). 07:14:01 -!- GregorR-L has joined. 07:14:03 http://www.cafepress.com/bizarregeek.255291006 07:14:55 (For those who like CAs :P ) 07:42:46 Deewiant, not that mine works perfectly either so far 07:43:08 Deewiant, it fails when using garbage collector for unknown reason 07:59:59 -!- clog has quit (ended). 08:00:00 -!- clog has joined. 08:10:35 -!- adu has quit ("Bye"). 08:34:12 -!- olsner has quit ("Leaving"). 08:36:00 -!- Iskr has joined. 09:02:45 Deewiant, actually the issue seems to depend on how you interpret the PERL fingerprint docs... hm... 09:03:29 "-e print 'A',eval(%s)" or "-e print 'A';eval(%s)" 09:03:34 where %s is the popped string 10:16:42 -!- GregorR-L has quit (Read error: 110 (Connection timed out)). 10:20:48 okokoko 10:20:51 -> 11:06:50 -!- timotiis has joined. 11:28:09 AnMaster: okay, so we want the Perl program to be able to use the std streams as normal, but how do we get the eval result out then? 11:28:36 I'm confused as to how that would be possible 11:32:35 would just "$x = eval(%s); print 'A',$x" work? can the %s mess with a variable called $x in some way? (I'm unaware of the limitations of eval in Perl) 11:43:23 -!- dbc has joined. 11:56:57 and hmm, even that wouldn't work, since %s might print A and $x might contain A 12:02:17 hm ok 12:02:37 Deewiant, I'm no perl expert either 12:02:40 far from 12:03:24 bbl 12:03:27 I'd say the simplest solution here is to say that accessing stdout is undefined 12:03:37 alternatively make it stderr for more convenience 12:07:49 -!- Sgeo has joined. 12:36:13 -!- jix has joined. 12:59:32 Deewiant, I capture both stdout and stderr 12:59:34 that isn't hard 13:00:28 AnMaster: yes, but how do you get the eval result 13:01:42 Deewiant, just int fds[2]; pipe(fds); pid = fork() if (pid == 0) { /* in child */ dup2(fds[1], 1); dup2(fds[1], 2); .... execvp("perl", arguments); } else { ... read(fds[0], ...) ... } 13:01:54 yes, capturing stderr is trivial 13:01:55 I know this 13:01:56 Deewiant, well a good question 13:02:16 the specs are unclear on what to exactly capture I think? 13:02:21 what exactly to* 13:02:24 no, I think it's clear 13:02:28 hm 13:02:33 it's the value of the expression evaled 13:02:41 hm 13:02:52 e.g. eval 2+2 and get 4. 13:02:54 what about loading libperl.so, I think you can do something like tht 13:02:56 that* 13:03:40 possibly... but it shouldn't be necessary 13:03:51 true 13:03:54 based on what the docs say, as well 13:04:05 anyway then S would get a lot more complex 13:04:09 "It pushes a 1 on the stack otherwise, indicating that the Perl language will be shelled when needed. 13:04:11 say "isn't loaded yet" 13:04:25 hm true 13:04:40 that speaks explicitly of shelling, so libperl.so is a no-go ;-) 13:04:58 I haven't read the whole discussion, but based on what I've seen: no-one says you need to limit yourself to STDOUT and STDERR, why not just open a pipe for file descriptor 3, and in Perl do open FD3, ">&3"; print FD3 eval(...); 13:05:11 anyway, now I really understand why ppl don't use read() but use fread() instead 13:05:27 it a lot easier to work with 13:05:54 fizzie, interesting 13:06:08 fizzie: isn't it possible for the perl code to always be using FD3 itself, though 13:06:26 fizzie, that would print both the result of the eval as well as any print statements inside the eval? 13:06:29 not very likely, but possible? 13:07:16 and hmm, I wonder if that's possible on windows 13:07:32 Well, you'd get stdout, stderr and that eval result via three different pipes. But sure, the Perl code could use file descriptor 3 for something. 13:07:46 hm 13:10:33 also I wonder why I get a segfault with boehm-gc if I use execlp after fork(), execvp works 13:10:49 it makes no sense to me 13:14:59 even more interesting is using valgrind (GC disabled then) 13:15:16 using --trace-children=yes cause things to fail if I use execvp, but they work with execlp 13:15:22 only under valgrind that is 13:17:36 Deewiant, btw, does windows have vfork()? 13:17:43 windows doesn't have fork 13:17:44 of any kind 13:18:08 so how do you start another process without replacing the current one 13:18:22 CreateProcessA/CreateProcessW 13:18:27 don't ask me about the details 13:18:35 I just use what Tango provides 13:19:07 as far as I can see you need to fork() then { in child: dup2 and exec } { in parent: waitpid and read } 13:19:14 hm 13:19:22 windows doesn't use the fork/exec model 13:19:30 I see 13:19:53 hm actually I spend most code at doing read and trying to handle all the error conditions/results from it 13:20:05 yep, same here 13:20:06 ie, "do we need to realloc and read some more or are we done yet" 13:20:09 just running the program is 2 lines :-P 13:20:24 well a bit more for building arguments and such here 13:20:36 but the read code is really a mess 13:22:13 I think doing read into same buffer and keeping extending it may be the wrong way, read and then append the buffer to another buffer would be way easier, and, due to \0 termination, faster 13:22:17 I think it'd be fair to just disallow stderr for PERL 13:22:26 there's no way to write to stderr in Befunge otherwise either 13:22:29 hm? well capturing it isn't an issue 13:22:38 no 13:22:43 but then you can reserve it for the eval result 13:22:51 anyway I use a perror() after forking 13:22:52 i.e. print stderr eval(%s) 13:23:08 in case of exec fails 13:23:22 and then you can pass the perl script's stdout through to the interpreter's stdout 13:23:31 whilst capturing the stderr and reading the eval result from it 13:23:52 Deewiant, hm and what about print stderr eval(print stdout "foo")? 13:23:58 hm 13:24:01 2008-04-16 15:23:21 ( Deewiant) and then you can pass the perl script's stdout through to the interpreter's stdout 13:24:09 Deewiant, not sure if that is a good thing or not 13:24:19 I think it's the least error prone option 13:24:26 + simple to implement :-P 13:25:43 because it allows the PERL script to do output if it so wishes 13:25:58 Deewiant, btw where do you close the fd in eval in perl.d 13:26:06 or is D handling that for you? 13:26:12 Tango does it 13:26:26 in p.execute(); I would guess 13:26:50 read = p.stdout.input.read(buf); 13:26:52 hm 13:26:57 stdout.input? 13:27:13 the input stream of the process's stdout 13:27:32 err, stdout is just output, it isn't bi-directional 13:27:50 but if another program is writing to it what are you going to do 13:27:54 you are going to read from it, no? 13:27:55 so it's input 13:28:08 similarly stdin is an output stream 13:28:09 hm... yeah 13:28:19 Deewiant, but then is stdout.output invalid? 13:28:28 when done on another process 13:28:32 yes, I think so 13:29:02 You could also in Perl do something like: open(REALERR, ">&STDERR); open(STDERR, ">&STDOUT); print REALERR eval(...); if you wanted to have all the eval'd piece outputs (both stdout and stderr) mixed in the stdout of the executed process, leaving the real stderr for the eval result. Of course the eval'd code could still get to the real stderr if it really wanted to. 13:29:10 Deewiant, what is the print A bit about really btw 13:29:25 AnMaster: it took me a while to figure out, too ;-) 13:29:40 Deewiant, well you do it, I don't see why thoug 13:29:41 AnMaster: the idea being that if %s prints to stdout, we take only the eval result. 13:29:41 though 13:29:46 I think. 13:29:49 ah I see 13:29:59 fizzie, heh 13:30:00 but of course that doesn't work if the eval result contains A so it's crap. :-P 13:30:26 fizzie, I know about 0 perl, I do know hello world in perl and I do know PCRE (by libpcre) 13:30:29 but that's about it 13:30:33 fizzie: hm, that might be a good idea actually. 13:30:59 fizzie, so forgive me if I'm ignorant, but the quotes seem mismatched? 13:31:12 open(REALERR, ">&STDERR); <-- as in that 13:31:24 Yes, there's two "s missing. I was about to metion it, but then didn't bother. Should be ">&STDERR". 13:31:30 thanks 13:31:51 fizzie: and then just tell the eval'd code not to mess with a stream called REALERR? Or is there another way it can subvert the whole thing? 13:32:28 Well, the eval'd code could use open(WHATEVER, ">&2"); and use WHATEVER to print to the real stderr file descriptor. 13:32:57 Darn. There's no way to hide the real stderr completely? 13:33:20 I don't think there is, except of course closing it, but that kind-of defeats the purpose. 13:34:04 Alright. Hmh. 13:34:58 Perl might have some sandbox-y functions (probably as CPAN modules) for limiting the things the eval'd code can do. And it might be possible to hook into Perl's "open" call behaviour somehow, but that would mean even more Perl-guts. 13:35:41 ugh. 13:36:59 hm 13:37:05 killall perl? ;) 13:38:02 fizzie, would calling it say, CFUNGE_REALERR work 13:38:17 then I could just say "your own fault, cfunge is my namespace" 13:38:19 AnMaster: unless the code uses something called CFUNGE_REALERR. 13:38:26 AnMaster: or opens ">&2". 13:38:53 point is, is _ valid or not 13:39:04 sure 13:39:17 well considering it's perl, it is bound to have some special meaning 13:39:18 if not, call it CFUNGEISMYNAMESPACEYOUDORKREALERR 13:39:18 :-P 13:39:28 it's a lot like C lexically. 13:39:56 well iirc it does have packages, ie use foo; not #include 13:40:08 thought they provided kind of C++ namespaces 13:40:14 I may totally wrong 13:40:38 yes, not 'the same as C', just 'like C'. 13:40:52 Actually you could open(my $realerr, ">&STDERR); open(...); eval(...); which I _think_ would mean that the eval'd code wouldn't be able to access the filehandle directly, since there's no symbol table entry for it, just the reference in $realerr. (Can't remember details about lexically scoped "my" vars and eval, but at least with some minor changes.) 13:41:11 Still won't help with the open(FOO, ">&2"); problem. 13:41:36 hm. 13:41:40 (With open my $realerr you'd then do print $realerr eval(...);) 13:42:14 wait a second, is $ same as in bash or something else? 13:42:30 $ is just the prefix for a scalar variable. 13:42:44 mhm same as in bash more or less then 13:42:56 fizzie: perl -e 'my $x = 2; print eval($x+$x);' prints 4, at least. 13:43:23 is that evaluated lexically like a pre-processor first? 13:43:49 -!- atsampson has quit (Read error: 113 (No route to host)). 13:43:54 in bash you would need to escape the \$x in an eval statement 13:44:08 because of order of expanding 13:44:41 (however I will not write a perl escaper in C or something like that) 13:47:20 Well, if nothing else works, you can always use print $realerr eval {$realerr = undef; %s}; but that's butt-ugly. I'm pretty sure there is a better way for making a variable that doesn't leak into eval; lexically scoped vars evidently do. 13:48:34 fizzie, there probably is indeed, it's perl after all 13:48:41 Actually more like eval { local $realerr = undef; %s } to avoid gobbling the real $realerr, but then it can't be a "my" variable. 13:48:59 it is probably something like z having a special meaning in a special context *runs* 13:49:09 remember, perl DOES have a entry on the esolang wiki 13:49:58 I think someone said that instead of "foo" you could do q/foo/ in perl. 13:50:02 or something like that 13:50:08 you can do that in D 2.0 as well. :-) 13:50:25 Deewiant, s/)/(/ 13:50:45 :-P 13:50:55 and that was sed not perl! 13:51:17 I have no problems using sed or awk 13:51:19 it's also perl. 13:51:29 Deewiant, yeah I thought so 13:51:31 Actually, if you're really doing string-substitution in the Perl code itself, you don't need eval, and this should work: my $code = sub { %s }; open my $realerr, ">&STDERR"; open STDERR, ">&STDOUT"; print $realerr $code->(); 13:51:45 fizzie, eh? 13:51:56 fizzie: unfortunately the specs say that it should be an eval(). :-) 13:51:57 "E ('Eval') pops a 0gnirts string and performs a Perl eval() on it, possibly (or not) shelling Perl as indicated by S above. The result of the call is pushed as a 0gnirts string back onto the stack." 13:52:07 fizzie, from http://catseye.tc/projects/funge98/library/PERL.html 13:53:11 Oh. Well, my $code = sub { return eval('%s'); } ... then. 13:53:44 in fact, isn't eval(%s) equivalent to %s always? 13:54:10 where %s isn't wrapped in "" or '' or whatever of course 13:54:13 Deewiant, hm... couldn't you pass a string to eval, read from say stdin? 13:54:50 yes, but I'm just pondering. 13:55:06 I don't think eval(%s) really does much, since the %s is evaluated normally, and the result is then passed to the 'eval' function. 13:55:26 eval(my $x = 1; print $x) versus my $x = 1; print $x 13:55:53 ah 13:55:55 "Alternatively, if given a block of code - that is, the code is known at compile time - eval traps run-time exceptions." 13:56:08 what does that mean Deewiant ? 13:56:14 Actually "eval(my $x = 1; print $x)" shouldn't be syntactically right at all. 13:56:16 divide by zero or? 13:56:18 I don't know 13:56:23 something like that probably 13:56:30 fizzie, it shouldn't? 13:56:31 fizzie: only expressions? 13:56:35 Eval takes either a string or a block of code and evaluates it. 13:56:45 hm 13:56:50 perl -e 'eval(my $x = 1; print $x);' gives syntax error at -e line 1, near "1;" 13:57:08 fizzie, perl -e 'eval(print "bar");' works though, hm 13:57:12 alright 13:57:27 AnMaster: that's an expression. you can pass the result of printf to a function in C just fine as well. 13:57:34 hm 13:57:39 right 13:58:10 Since the spec speaks of strings, I think it means one should do something like eval('%s'); but that would need escaping the 's in the string. 13:58:16 (Or reading it from stdin or some-such.) 13:58:31 um 13:58:34 perl -e 'print eval(4 + 5)' 13:58:42 is not same as perl -e '4 + 5' 13:58:53 and with multistatement just adding a print doesn't work 13:59:43 you said something about code block? 14:00:34 and then there is the question of non-matching ( and ) in the input hm... 14:00:41 perl -e 'print eval("my $x = 1;")' prints nothing 14:00:49 but it doesn't error out either 14:01:12 "print eval(%s)", "10); something here(blah" 14:01:18 yay perl injection? 14:01:22 perl -e 'print eval("my \$x = 1;")' however prints "1". 14:01:49 argh, the shell goes inside "" inside '' 14:02:06 stupid bash, just do what I mean 14:02:08 Deewiant, err? bash doesn't at least 14:02:17 well evidently it does 14:02:19 perl -e 'print eval("my \$x = 1;")' works here 14:02:25 I think it's actually the Perl that does mess the $x inside a ""-string. 14:02:27 or no, is that perl mucking it up 14:02:34 Deewiant, indeed it is perl 14:02:35 ayeh 14:02:39 makes sense 14:02:54 fizzie, so given the specs, how would you do it? 14:03:09 I think sub { return eval('%s'); } is the best idea so far 14:03:22 need to do escaping in %s though 14:03:26 which is tricky 14:03:31 yeah something I don't want to mess with 14:03:45 considering the syntax of perl probably could escape escaping 14:03:48 so, instead, return eval(%s); 14:03:58 and statements are disallowed thus. 14:04:08 hm doesn't that break the specs? 14:04:17 But eval(%s) makes really no sense. The '%s' part is evaluated before even passing the result to 'eval'. 14:04:17 how? 14:04:37 fizzie, hm ok 14:04:47 then the specs are confused or PERL is practically impossible to implement without writing the interpreter in perl. 14:05:06 fizzie, could you read the stuff for eval from stdin? 14:05:18 because then perl would escape it itself kind of 14:05:33 Well, you could. Then you'd just use eval($s); after reading into $s. 14:05:42 ie, $variable = read(); eval($variable) 14:05:49 I don't know the needed syntax 14:06:03 fizzie, and how would you read that? 14:06:10 AnMaster: and if the perl script wants to read stdin? 14:06:17 Deewiant, hm true 14:06:25 Well, you'd probably need to have some sort of length-prefix before the code, so that you can keep the stdin open for further reads. 14:06:38 aye. 14:06:56 also it need to be blocking read 14:07:15 so if it get to read before the parent got to write, it would wait 14:08:24 fizzie, well if you can write the needed perl code to read like that, I could do the C bit 14:08:33 I wouldn't be able to do the perl bit sadly :/ 14:08:39 perl is trivial to learn 14:08:47 for short stuff like that I mean 14:08:49 If you don't want to use a prefix, you could use that file descriptor 3, read the code-to-evaluate and then close it. 14:08:57 fizzie, prefix could work 14:09:03 prefix is fine 14:09:10 I guess just write(fd, strlen(perlcode))? 14:09:20 however you choose to implement it 14:09:28 I guess that'd work the easiest 14:09:30 or if you want as a string 14:09:43 Well, it's easier to "read until eof" in one line of Perl than it is to "read at least this many bytes". 14:09:50 ie snprintf("%d", strlen...") 14:10:07 hm EOF isn't a special char is it 14:10:09 it's easier in most languages without scanf. 14:10:15 EOF isn't a char. 14:10:25 or even with scanf, I guess. 14:10:55 fizzie, well, input would be a pipe created with pipe() in C in my case 14:10:55 my $code = ''; open FD3, "<&3"; $code .= $_ while ; close FD3; might do it. 14:11:12 using just stdin makes that a bit easier. 14:11:16 not sure how I would send a EOF that way 14:11:23 my $code = ''; $code .= $_ while <>; 14:11:29 or what 14:11:37 AnMaster: you send EOF by closing the pipe. 14:11:42 or file descriptor, or whatever. 14:11:45 hm ok 14:11:57 Deewiant, well a pipe is a set of two connected fd's in this case 14:12:14 If you close the write end, the read end should get an eof. 14:12:17 well you close the input stream. 14:16:32 fizzie, then print eval($code) or? 14:16:39 or "$code"? 14:17:25 no difference in perl, just like in bash :-) 14:17:32 Yep. 14:17:43 Deewiant, there is a difference in bash 14:17:49 foo="bar quux" 14:17:53 echo $foo 14:17:55 nope, no difference. 14:17:55 echo "$foo" 14:18:01 ? 14:18:03 the difference is that $foo will expand to two parameters 14:18:09 yes 14:18:10 while "$foo" won't 14:18:21 for echo it doesn't matter 14:18:23 but the string itself is the same, it's just the way it's passed to another program that differs 14:18:28 for say grep on a filename it would 14:18:33 -!- timotiis has quit (Connection timed out). 14:18:33 if the filename contains spaces 14:18:38 since eval() is a builtin function, not another program, it makes no difference. 14:18:49 I would assume that it's the same for any builtin function of bash. 14:18:51 Deewiant, for bash it matters for builtins too 14:18:59 example? O_o 14:19:35 echo will do the same, but say: foo="echo bar" 14:19:36 then 14:19:40 $foo 14:19:43 and "$foo" 14:19:47 would be different 14:19:50 the first would echo bar 14:19:51 -!- timotiis has joined. 14:20:06 the second would likely say "echo bar" not such command or something like that 14:20:36 Deewiant, and say test -f $file and test -f "$file" would differ when $file contains spaces 14:20:53 test is a builtin yes 14:21:33 Deewiant, oh and: foo="bar quux"; for i in $foo; do echo "$i"; done 14:21:43 is different from: foo="bar quux"; for i in "$foo"; do echo "$i"; done 14:21:50 Bash is indeed prone to splitting things at spaces; but in Perl I think there's a difference only if $foo doesn't happen to be a string. 14:22:16 fizzie, well there are exceptions: 14:22:23 foo="abc def" 14:22:26 bar=$foo 14:22:33 would be the same as: 14:22:35 bar="$foo" 14:22:37 in that case 14:22:40 however... 14:22:50 foo="abc def" is not the same as foo=abc def 14:23:05 the latter would try to run def with foo=abc in the environment 14:23:15 Deewiant, enough examples? 14:23:17 ;P 14:24:33 fizzie, would this mean something like: 14:24:45 my $code = ''; $code .= $_ while <>; open(CFUNGE_REALERR, ">&STDERR"); open(STDERR, ">&STDOUT"); print CFUNGE_REALERR eval($code) 14:24:45 ? 14:25:39 Yes, if you don't mind losing the use of STDIN in the Perl code. (And of course you can avoid the CFUNGE_REALERR "problem" with a bit more trickery.) 14:26:26 well how would I then save and restore the stdin? 14:26:38 or actually I could just use fd3 I guess 14:26:46 or just pass the length first. 14:26:54 and then read only bytes to $code. 14:27:09 Yes, but with the length-prefix the reading is more complicated. (Of course with fd3 the writing is a bit more complicated.) 14:27:20 "my $code = ''; $code .= $_ while ; close FD3; open(CFUNGE_REALERR, \">&STDERR\"); open(STDERR, \">&STDOUT\"); print CFUNGE_REALERR eval($code)" 14:27:33