From steve at devon-it.co.uk  Thu Jul 20 05:23:23 2006
From: steve at devon-it.co.uk (Steve Marvell)
Date: Thu, 20 Jul 2006 13:23:23 +0100
Subject: [DCPM] windows newlines in #! line
Message-ID: <20060720122323.GA31115@devon-it.co.uk>

This started as a CGI problem. Apparently, some of the Perl scripts a
particular company provide are littered with Windows newlines. This
casuse a bad interpretor error.

This is also the case if the script is run by hand.

I understand that the present version of a Apache the clinet is using
handles this. 

I take this is a shell problem, but I think it could be solved if
Apache prestripped them

Anyone got any ideas?

Steve

From simon at technocool.net  Thu Jul 20 12:07:42 2006
From: simon at technocool.net (Simon Waters)
Date: Thu, 20 Jul 2006 20:07:42 +0100
Subject: [DCPM] windows newlines in #! line
In-Reply-To: <20060720122323.GA31115@devon-it.co.uk>
References: <20060720122323.GA31115@devon-it.co.uk>
Message-ID: <44BFD47E.7010401@technocool.net>

Steve Marvell wrote:
> This started as a CGI problem. Apparently, some of the Perl scripts a
> particular company provide are littered with Windows newlines. This
> casuse a bad interpretor error.
> 
> This is also the case if the script is run by hand.
> 
> I understand that the present version of a Apache the clinet is using
> handles this. 
> 
> I take this is a shell problem, but I think it could be solved if
> Apache prestripped them
> 
> Anyone got any ideas?

The usual cause is incorrectly transferring the files to the server. ftp
will correctly translate text files, but if they are shipped up as a zip
file then the transfer is skipped.

Just;

find . -name "*.pl" -print0 | xargs -0 dos2unix

Or whatever appropriate pattern is required for the find.

From user_groups at pm.org  Thu Jul 20 12:24:04 2006
From: user_groups at pm.org (Dave Cross)
Date: Thu, 20 Jul 2006 20:24:04 +0100
Subject: [DCPM] windows newlines in #! line
In-Reply-To: <20060720122323.GA31115@devon-it.co.uk>
References: <20060720122323.GA31115@devon-it.co.uk>
Message-ID: <44BFD854.70200@pm.org>

Steve Marvell wrote:
> This started as a CGI problem. Apparently, some of the Perl scripts a
> particular company provide are littered with Windows newlines. This
> casuse a bad interpretor error.
> 
> This is also the case if the script is run by hand.
> 
> I understand that the present version of a Apache the clinet is using
> handles this. 
> 
> I take this is a shell problem, but I think it could be solved if
> Apache prestripped them
> 
> Anyone got any ideas?

The problem is that your shebang line has a Windows newline character on 
the end. It therefore looks something like this:

   #!/usr/bin/perl\r\n

Unix recognises the \n as the end of the line and therefore looks for a 
program called

   /usr/bin/perl\r

And (usually) that doesn't exist.

Of course, you shouldn't be seeing this problem as your shebang line 
should look like this:

   #/usr/bin/perl -T\r\n

In this case, the shell correctly extracts the path to Perl and passes 
the slight broken command line argument (-T\r) to Perl. Perl can deal 
with broken line endings just fine, so this will fix the problem.

Of course, you'll then find that your program isn't taint-clean and 
that'll give you some far more serious problems to fix :)

Dave...


From steve at devon-it.co.uk  Fri Jul 21 03:50:22 2006
From: steve at devon-it.co.uk (Steve Marvell)
Date: Fri, 21 Jul 2006 11:50:22 +0100
Subject: [DCPM] windows newlines in #! line
In-Reply-To: <44BFD47E.7010401@technocool.net>
References: <20060720122323.GA31115@devon-it.co.uk>
	<44BFD47E.7010401@technocool.net>
Message-ID: <20060721105021.GB24782@devon-it.co.uk>

Simon Waters wrote:

> The usual cause is incorrectly transferring the files to the server. ftp
> will correctly translate text files,

FTP server or client?

> find . -name "*.pl" -print0 | xargs -0 dos2unix

That's good for me, but not good for downstream box users who are just
web clients.

Steve

From dave at dave.org.uk  Sat Jul 22 13:52:36 2006
From: dave at dave.org.uk (Dave Cross)
Date: Sat, 22 Jul 2006 21:52:36 +0100
Subject: [DCPM] windows newlines in #! line
In-Reply-To: <20060720122323.GA31115@devon-it.co.uk>
References: <20060720122323.GA31115@devon-it.co.uk>
Message-ID: <44C29014.10000@dave.org.uk>

Steve Marvell wrote:
> This started as a CGI problem. Apparently, some of the Perl scripts a
> particular company provide are littered with Windows newlines. This
> casuse a bad interpretor error.
> 
> This is also the case if the script is run by hand.
> 
> I understand that the present version of a Apache the clinet is using
> handles this. 
> 
> I take this is a shell problem, but I think it could be solved if
> Apache prestripped them
> 
> Anyone got any ideas?

The problem is that your shebang line has a Windows newline character on
the end. It therefore looks something like this:

   #!/usr/bin/perl\r\n

Unix recognises the \n as the end of the line and therefore looks for a
program called

   /usr/bin/perl\r

And (usually) that doesn't exist.

Of course, you shouldn't be seeing this problem as your shebang line
should look like this:

   #/usr/bin/perl -T\r\n

In this case, the shell correctly extracts the path to Perl and passes
the slight broken command line argument (-T\r) to Perl. Perl can deal
with broken line endings just fine, so this will fix the problem.

Of course, you'll then find that your program isn't taint-clean and
that'll give you some far more serious problems to fix :)

Dave...