[9fans] sed crash

[9fans] sed crash

[Uriel]

For those that enjoy fixing bugs:

% echo foo | sed
's/..a|..b|..c|..d|..e|..f|..g|..h|..i|..j|..k|..l|..m|..n|..o|..p//g'

Will crash both native and p9p sed.

Yes, I know it is a nonsensical expression, but it is the simplest one
that still made it crash, I have lost the original version which was
much more complex.

Note that providing two characters as input will still crash, but if
the input line is a single character it wont crash.

I have been meaning to get all the fun of tracking this one down
myself for a while, but seems that I never get to it, so I will let
others enjoy the fun of fixing it.

Peace

uriel

Re: [9fans] sed crash

[Pietro Gagliardi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Feb 4, 2009, at 9:28 PM, Uriel wrote:

> For those that enjoy fixing bugs:
>
> % echo foo | sed
> 's/..a|..b|..c|..d|..e|..f|..g|..h|..i|..j|..k|..l|..m|..n|..o|..p//g'
>
> Will crash both native and p9p sed.
>
> Yes, I know it is a nonsensical expression, but it is the simplest one
> that still made it crash, I have lost the original version which was
> much more complex.
>
> Note that providing two characters as input will still crash, but if
> the input line is a single character it wont crash.
>
> I have been meaning to get all the fun of tracking this one down
> myself for a while, but seems that I never get to it, so I will let
> others enjoy the fun of fixing it.
>
> Peace
>
> uriel
>
For me, it crashes on the second line of input, in the dosub()
function. Will investigate now.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEYEARECAAYFAkmKVgoACgkQuv7AVNQDs+x7ZACfYvtxSB5c8TD8LFFjPPHsgtah
9DwAnjIQ0GM4b82JGPpdJIUYfHjs7+Pv
=7lRk
-----END PGP SIGNATURE-----

Re: [9fans] sed crash

[Pietro Gagliardi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

	lp = loc2;
	loc2 = sp - genbuf + linebuf;
	while(*sp++ = *lp++)

At the point of crash, loc2 is a null pointer when it is assigned to
lp, so when the while() statement executes, crash. Now to find out why
and when loc2 becomes null.

On Feb 4, 2009, at 9:59 PM, Pietro Gagliardi wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On Feb 4, 2009, at 9:28 PM, Uriel wrote:
>
>> For those that enjoy fixing bugs:
>>
>> % echo foo | sed
>> 's/..a|..b|..c|..d|..e|..f|..g|..h|..i|..j|..k|..l|..m|..n|..o|..p//
>> g'
>>
>> Will crash both native and p9p sed.
>>
>> Yes, I know it is a nonsensical expression, but it is the simplest
>> one
>> that still made it crash, I have lost the original version which was
>> much more complex.
>>
>> Note that providing two characters as input will still crash, but if
>> the input line is a single character it wont crash.
>>
>> I have been meaning to get all the fun of tracking this one down
>> myself for a while, but seems that I never get to it, so I will let
>> others enjoy the fun of fixing it.
>>
>> Peace
>>
>> uriel
>>
> For me, it crashes on the second line of input, in the dosub()
> function. Will investigate now.
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.8 (Darwin)
>
> iEYEARECAAYFAkmKVgoACgkQuv7AVNQDs+x7ZACfYvtxSB5c8TD8LFFjPPHsgtah
> 9DwAnjIQ0GM4b82JGPpdJIUYfHjs7+Pv
> =7lRk
> -----END PGP SIGNATURE-----
>

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEYEARECAAYFAkmKWCEACgkQuv7AVNQDs+xSQQCgoFPwqMsaphtaKBTHHLFryiJr
eUMAnA9BjuDpMU40VqJp5hu1EDr3mdXp
=XF/E
-----END PGP SIGNATURE-----

Re: [9fans] sed crash

[Pietro Gagliardi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The bug doesn't seem to be in sed. The match() function calls
rregexec(), which sets a strucuture that is set to loc2, which is made
null for some reason.

On Feb 4, 2009, at 10:08 PM, Pietro Gagliardi wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> 	lp = loc2;
> 	loc2 = sp - genbuf + linebuf;
> 	while(*sp++ = *lp++)
>
> At the point of crash, loc2 is a null pointer when it is assigned to
> lp, so when the while() statement executes, crash. Now to find out
> why and when loc2 becomes null.
>
> On Feb 4, 2009, at 9:59 PM, Pietro Gagliardi wrote:
>
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> On Feb 4, 2009, at 9:28 PM, Uriel wrote:
>>
>>> For those that enjoy fixing bugs:
>>>
>>> % echo foo | sed
>>> 's
>>> /..a|..b|..c|..d|..e|..f|..g|..h|..i|..j|..k|..l|..m|..n|..o|..p//g'
>>>
>>> Will crash both native and p9p sed.
>>>
>>> Yes, I know it is a nonsensical expression, but it is the simplest
>>> one
>>> that still made it crash, I have lost the original version which was
>>> much more complex.
>>>
>>> Note that providing two characters as input will still crash, but if
>>> the input line is a single character it wont crash.
>>>
>>> I have been meaning to get all the fun of tracking this one down
>>> myself for a while, but seems that I never get to it, so I will let
>>> others enjoy the fun of fixing it.
>>>
>>> Peace
>>>
>>> uriel
>>>
>> For me, it crashes on the second line of input, in the dosub()
>> function. Will investigate now.
>> -----BEGIN PGP SIGNATURE-----
>> Version: GnuPG v1.4.8 (Darwin)
>>
>> iEYEARECAAYFAkmKVgoACgkQuv7AVNQDs+x7ZACfYvtxSB5c8TD8LFFjPPHsgtah
>> 9DwAnjIQ0GM4b82JGPpdJIUYfHjs7+Pv
>> =7lRk
>> -----END PGP SIGNATURE-----
>>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.8 (Darwin)
>
> iEYEARECAAYFAkmKWCEACgkQuv7AVNQDs+xSQQCgoFPwqMsaphtaKBTHHLFryiJr
> eUMAnA9BjuDpMU40VqJp5hu1EDr3mdXp
> =XF/E
> -----END PGP SIGNATURE-----
>

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEUEARECAAYFAkmKYMwACgkQuv7AVNQDs+xcpACYi9wXYM/0ONUOOrqPJu4ViRFZ
BACfTLv0GRuZzUP/WYeU72Jm+KqDunw=
=Vrfb
-----END PGP SIGNATURE-----

Re: [9fans] sed crash

[Akshat Kumar]

Dear Pietro,
Could you please remove the PGP Signatures from
your messages?


don't do unto us, illegibly
ak

Re: [9fans] sed crash

[andrey mirtchovski]

also, can you please remove debugging output from 9fans?

Re: [9fans] sed crash

[Russ Cox]

Sed is fixed in plan9port.

$ hg diff sed.c
diff -r 478b53880a80 src/cmd/sed.c
--- a/src/cmd/sed.c	Wed Feb 04 20:12:48 2009 -0800
+++ b/src/cmd/sed.c	Wed Feb 04 20:13:08 2009 -0800
@@ -994,7 +994,7 @@
 		return 0;
 	subexp[0].s.rsp = buf;
 	subexp[0].e.rep = 0;
-	if (rregexec(pattern, linebuf, subexp, MAXSUB)) {
+	if (rregexec(pattern, linebuf, subexp, MAXSUB) > 0) {
 		loc1 = subexp[0].s.rsp;
 		loc2 = subexp[0].e.rep;
 		return 1;
$

http://hg.pdos.csail.mit.edu/hg/plan9/rev/0bf86c524c9b

Re: [9fans] sed crash

[Pietro Gagliardi]

On Feb 4, 2009, at 10:58 PM, andrey mirtchovski wrote:

> also, can you please remove debugging output from 9fans?

What do you mean? Could you paste the raw text data for one of my
emails? PGP for Mac Mail hides everything.

Re: [9fans] sed crash

[Iruata Souza]

On Thu, Feb 5, 2009 at 2:39 AM, Pietro Gagliardi <pietro10@mac.com> wrote:
> On Feb 4, 2009, at 10:58 PM, andrey mirtchovski wrote:
>
>> also, can you please remove debugging output from 9fans?
>
> What do you mean? Could you paste the raw text data for one of my emails?
> PGP for Mac Mail hides everything.
>
>

maybe he's trying to say this is a mailing list, not twitter.
correct me if i'm wrong, andrey.

iru

Re: [9fans] sed crash

[erik quanstrom]

>
> What do you mean? Could you paste the raw text data for one of my
> emails? PGP for Mac Mail hides everything.
>

i considered adding code to upas to at least
strip pgp nonesense from message bodies.  but this
would require upas to know things about what
message bodies look like after mime is done with
them.  i thought that was a bad idea.  next we'll
be replacing :-) (':' '-' ')' for those with impaired
mail readers) with a jpeg.

- erik

Re: [9fans] sed crash

[Joel C. Salomon]

erik quanstrom wrote:
> next we'll be replacing :-) (':' '-' ')' for those with impaired mail
readers) with a jpeg.

This is Plan 9: replace it with ☺.

—Joel

Re: [9fans] sed crash

[Uriel]

That was even faster than I expected. Thanks Russ!

uriel

P.S.: Silly question, why is du(1) not built by default in p9p? The
code already there seems to work as far as I can tell..

On Thu, Feb 5, 2009 at 5:15 AM, Russ Cox <rsc@swtch.com> wrote:
> Sed is fixed in plan9port.
>
> $ hg diff sed.c
> diff -r 478b53880a80 src/cmd/sed.c
> --- a/src/cmd/sed.c     Wed Feb 04 20:12:48 2009 -0800
> +++ b/src/cmd/sed.c     Wed Feb 04 20:13:08 2009 -0800
> @@ -994,7 +994,7 @@
>                return 0;
>        subexp[0].s.rsp = buf;
>        subexp[0].e.rep = 0;
> -       if (rregexec(pattern, linebuf, subexp, MAXSUB)) {
> +       if (rregexec(pattern, linebuf, subexp, MAXSUB) > 0) {
>                loc1 = subexp[0].s.rsp;
>                loc2 = subexp[0].e.rep;
>                return 1;
> $
>
> http://hg.pdos.csail.mit.edu/hg/plan9/rev/0bf86c524c9b
>
>

Re: [9fans] sed crash

[erik quanstrom]

> erik quanstrom wrote:
> > next we'll be replacing :-) (':' '-' ')' for those with impaired mail
> readers) with a jpeg.
>
> This is Plan 9: replace it with ☺.

clearly.  " tcs -f emoticon" anyone? ☺

the point i'm trying to make is that "☺" is
not the same as :-).  i find it irritating that
some mail readers massage and process the
text to the point where one can see no
difference.

the reason this is irritating is that the mail
reader has decided that message bodies are
not text but rather an input language to it's
own special mail frobnicator.  since nobody
has admitted that this is a language, there
is no spec and i'd wager that no two mail
readers do quite the same thing.

- erik

Re: [9fans] sed crash

[Russ Cox]

> P.S.: Silly question, why is du(1) not built by default in p9p? The
> code already there seems to work as far as I can tell..

It's a long story, dating back to when the standard mode of
operation was to put $PLAN9/bin ahead of the system binaries
in your $PATH.  In that context it didn't make sense to include
tools like du, dd, and grep, because too many programs
expected to get the standard tool rather than the Plan 9 one.

The 9 script made it easier to live with $PLAN9/bin at the
end of the $PATH, so now there's no harm in having those.
I've re-enabled dd and du.

Russ

Re: [9fans] sed crash

[Pietro Gagliardi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I misunderstood what they meant by "debugging." PGP is doing nothing
wrong; I thought there was debugging info in the email.

On Feb 5, 2009, at 8:16 AM, erik quanstrom wrote:

>>
>> What do you mean? Could you paste the raw text data for one of my
>> emails? PGP for Mac Mail hides everything.
>>
>
> i considered adding code to upas to at least
> strip pgp nonesense from message bodies.  but this
> would require upas to know things about what
> message bodies look like after mime is done with
> them.  i thought that was a bad idea.  next we'll
> be replacing :-) (':' '-' ')' for those with impaired
> mail readers) with a jpeg.
>
> - erik
>

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (Darwin)

iEYEARECAAYFAkmLU9gACgkQuv7AVNQDs+xFiwCeJQOQC4pVv8sR5gFkEilI2tSP
O/gAoJNbwL5A0nBgd4182tT6GFXTsf+3
=hmuJ
-----END PGP SIGNATURE-----

Re: [9fans] sed crash

[erik quanstrom]

> I misunderstood what they meant by "debugging." PGP is doing nothing
> wrong; I thought there was debugging info in the email.

imho, it is against the spirit of MIME to put
encoded data in a text/plain body.  your
other post points out why funky non-standard
encodings are a problem.

it would be much preferable to put the pgp
stuff in its own mime part, as many mail readers
do.

- erik

Re: [9fans] sed crash

[john]

> I misunderstood what they meant by "debugging." PGP is doing nothing
> wrong; I thought there was debugging info in the email.
>
> On Feb 5, 2009, at 8:16 AM, erik quanstrom wrote:
>
>>>
>>> What do you mean? Could you paste the raw text data for one of my
>>> emails? PGP for Mac Mail hides everything.
>>>
>>
>> i considered adding code to upas to at least
>> strip pgp nonesense from message bodies.  but this
>> would require upas to know things about what
>> message bodies look like after mime is done with
>> them.  i thought that was a bad idea.  next we'll
>> be replacing :-) (':' '-' ')' for those with impaired
>> mail readers) with a jpeg.
>>
>> - erik
>>

The PGP doesn't bother me much; at least it's much more informative
and useful than "Hey, I know how to use acid to find out where a
process is dying!"

John

Re: [9fans] sed crash

[Uriel]

Cool, thanks again!

uriel

On Thu, Feb 5, 2009 at 6:27 PM, Russ Cox <rsc@swtch.com> wrote:
>> P.S.: Silly question, why is du(1) not built by default in p9p? The
>> code already there seems to work as far as I can tell..
>
> It's a long story, dating back to when the standard mode of
> operation was to put $PLAN9/bin ahead of the system binaries
> in your $PATH.  In that context it didn't make sense to include
> tools like du, dd, and grep, because too many programs
> expected to get the standard tool rather than the Plan 9 one.
>
> The 9 script made it easier to live with $PLAN9/bin at the
> end of the $PATH, so now there's no harm in having those.
> I've re-enabled dd and du.
>
> Russ
>
>

Re: [9fans] sed crash

[Akshat Kumar]

2009/2/4 Pietro Gagliardi <pietro10@mac.com>:
> PGP for Mac Mail hides everything.

upas and GMail have this capability also, but you might
find it more favourable to simply consider others on the
list, as well as the worth of the information you decide
to include.


filter:
from: pietro10@mac.com
*Delete*
ak

Re: [9fans] sed crash

[hiro]

>> PGP for Mac Mail hides everything.
>
> upas and GMail have this capability also, but you might

No, I ecperience the same annoying PGP SIGNATURE stuff on gmail.

Re: [9fans] sed crash

[Akshat Kumar]

2009/2/5 hiro <23hiro@googlemail.com>:
>>> PGP for Mac Mail hides everything.
>>
>> upas and GMail have this capability also, but you might
>
> No, I ecperience the same annoying PGP SIGNATURE stuff on gmail.
>
>

hiding falls out naturally from automated filtration/deletion

Re: [9fans] sed crash

[Derek Fawcus]

On Thu, Feb 05, 2009 at 04:13:46PM -0500, erik quanstrom wrote:
> it would be much preferable to put the pgp
> stuff in its own mime part, as many mail readers
> do.

  X-Pgp-Agent: GPGMail d55 (v55, Leopard)
  X-Mailer: Apple Mail (2.930.3)

and his mailer (plugin) can do so - if it is configured correctly.