From: jmichel AT schur DOT institut DOT math DOT jussieu DOT fr (Jean Michel)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: Bug in djgpp libc
Date: 1 Jan 2001 12:53:30 +0100
Organization: A poorly-installed InterNetNews site
Lines: 54
Message-ID: <92pr3q$12a$1@schur.institut.math.jussieu.fr>
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1010101101339 DOT 3587Z-100000 AT is>
NNTP-Posting-Host: schur.institut.math.jussieu.fr
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
X-Trace: vishnu.jussieu.fr 978349627 1036 134.157.13.71 (1 Jan 2001 11:47:07 GMT)
X-Complaints-To: Newsmaster AT jussieu DOT fr.
NNTP-Posting-Date: 1 Jan 2001 11:47:07 GMT
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

In article <Pine DOT SUN DOT 3 DOT 91 DOT 1010101101339 DOT 3587Z-100000 AT is>,
Eli Zaretskii  <djgpp AT delorie DOT com> wrote:
>
>On 31 Dec 2000, Jean Michel wrote:
>
>> Hello, there is a bug in the djgpp libc library.
>
>It's not a bug in DJGPP.
>
>> { printf("\032\n");
>>   fprintf(stderr, "%s\n", strerror(errno));
>> }
>> 
>> produces
>> 
>> No space left on drive (ENOSPC)
>
>That's because the DOS/Windows console driver returns EOF when you write 
>the Ctrl-Z character to it in cooked mode.  Ctrl-Z is a kind of 
>``software end-of-file'' marker in DOS.  There's nothing DJGPP can do with 
>that.

Perhaps return a more accurate (unique?) error condition?

>> [For people who may wonder why one would want to do the above, this is
>> the end result of trying to find the smallest example which shows a bug
>> occuring in the djgpp port of the ruby interpreter]
>
>A program which really _needs_ to write a raw Ctrl-Z character to the 
>screen, should switch the console device into raw mode, where the special 
>meaning of characters such as Ctrl-Z, Ctrl-P, TAB, etc. is switched off.
>
>However, since this issue is tricky, I'd suggest to post here an 
>explanation of why does the ruby interpreter need to write a Ctrl-Z 
>character to the screen.  It's quite possible that switching the console 
>into raw mode is not the right thing to do in that case, but it's hard to 
>suggest a good solution without knowing the details.

The fact is that  the interpreter does not know to  which file handle it
is writing to at  that stage, it is just echoing  characters read from a
text file which happens to contain the above (probably after having been
modified by some strange DOS editor). Actually the above code works fine
(the right  characters are written to  stdout). The only problem  is the
error return, together  with the decision in ruby to  raise an exception
whenever that  condition is reported  (a reasonable  thing to do  if you
have a real  ENOSPC). Since in C  one usually would not  check the error
return in the above  code, it tends to run fine. Again  I think the best
thing would be to invent some new errno (like errno=39); or realize that
things went fine  despite the error message... Also I  point out that no
error message is returned  if the ^Z is not followed very  soon by a \n.
Other characters after the ^Z cause no problem....

Best regards,
   Jean MICHEL