Re^5: ActiveState woes : Is it EOF-blind?

Actually, newer versions of perl have broken the 'proper' behaviour of text mode on Win32. Relative to the system utilities anyway. Historically, ^Z was used to indicate the end of file in text files whose content failed to fill the last allocation block on disc on filesystems that didn't have room in their directory structures for a number to determine this information. And whilst the reason for using ^Z has fallen into history, the system utilities still honour this mechanism(*).

If you create a text file that contains a ^Z part way through:

C:\test>perl -wle"
    print 'the quick brown fox' for 1..3; 
    print chr( 26 ); 
    print 'secret message'
" > junk.txt
[download]

And then use a system utility to display it:

C:\test>type junk.txt
the quick brown fox
the quick brown fox
the quick brown fox
[download]

The system utitlties stop reading when they encounter the ^Z, just like in days of yore.

Perl used to honour this in 5.6.1 days, but has since broken the behaviour:

C:\test>perl -pe1 junk.txt
the quick brown fox
the quick brown fox
the quick brown fox
&#8594;
secret message
[download]

translation to unicode entity the result of posting of course

Perl shouldn't read past a ^Z unless the file has been binmode'd. And yes, a consquence of this is that when reading unicode files, they should be read in binary mode.

I don't suppose that a 'bug' report for this would rate much attention. If the world has survived the 5(more?) years of 5.8.x without noticing this, then I don't expect much action to be taken now. But can you imagine the storm if Perl for *nix suddenly started to doing line-end translation unless you used binmode :)

(*)As a historical aside, when Mike Cowlishaw ported REXX to OS/2, he used this feature to good effect. The first time a REXX program was run, the 'compiled' binary bytecode was attached to the end of the file after the ^Z. On subsequent runs, if the bytecode was still there, it would be used rather than recompiling. As all the editors and system utilities used the MS CRT to read the files, whenever a file was copied or editied, the bytecode was transparently discarded and so it would be re-compiled whenever it was modified.

Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.

"Science is about questioning the status quo. Questioning authority".

In the absence of evidence, opinion is indistinguishable from prejudice.

"Too many [] have been sedated by an oppressive environment of political correctness and risk aversion."

Comment on Re^5: ActiveState woes : Is it EOF-blind? Select or Download Code

Replies are listed 'Best First'.
Re^6: ActiveState woes : Is it EOF-blind? by ikegami (Patriarch) on May 06, 2008 at 12:26 UTC
And whilst the reason for using ^Z has fallen into history, the system utilities still honour this mechanism Most do. For example, `type` currently stops at ^Z while `more` does not. In my experience, fewer and fewer programs supported it as times went by up until DOS 5 days (or so) where it became impossible to rely on. Regarding your asterisk, text editors were the first applications I noticed to stop honoring ^Z, so his trick would have been the first to fail for me.	[reply] [d/l] [select]

Replies are listed 'Best First'.

Re^6: ActiveState woes : Is it EOF-blind?
by ikegami (Patriarch) on May 06, 2008 at 12:26 UTC

And whilst the reason for using ^Z has fallen into history, the system utilities still honour this mechanism

*Most* do. For example, type currently stops at ^Z while more does not.

In my experience, fewer and fewer programs supported it as times went by up until DOS 5 days (or so) where it became impossible to rely on.

Regarding your asterisk, text editors were the first applications I noticed to stop honoring ^Z, so his trick would have been the first to fail for me.

[reply]
[d/l]
[select]