fseek

[Chris Torek]

There could be systems that keep text and binary files completely
separate, such that no operations are allowed on the wrong type.
I don't know of any, though.

VMS does something arguably more useful than that. In VMS, files
have a great deal of structure, which is normally managed by RMS
(the "record management system" -- code that, as Ken Thompson wrote
so long ago, "fills a much-needed gap" ). If you have a file
with (say) variable-length records, and open it as a fixed-length
record file, RMS will pad and (I presume) truncate records as
required to fit the data into the Procrustean bed. Similarly, if
you open a variable-length record file as text, RMS (or the C
runtime library -- pre-VMS-5.0 and Stream-LF files, I imagine it
was done outside RMS proper) can break up the records with
artificially-inserted newlines.

In any case, if one has ever actually used these operating systems
that fail to provide the much-needed gap -- i.e., the ones that
have a plethora of forcibly-imposed file formats -- one should
begin to understand the strong constraints put on the C programmer
by the C standards.

 

[glen herrmannsfeldt]

VMS does something arguably more useful than that. In VMS, files
have a great deal of structure, which is normally managed by RMS
(the "record management system" -- code that, as Ken Thompson wrote
so long ago, "fills a much-needed gap" ). If you have a file
with (say) variable-length records, and open it as a fixed-length
record file, RMS will pad and (I presume) truncate records as
required to fit the data into the Procrustean bed. Similarly, if
you open a variable-length record file as text, RMS (or the C
runtime library -- pre-VMS-5.0 and Stream-LF files, I imagine it
was done outside RMS proper) can break up the records with
artificially-inserted newlines.

(snip)

That sounds somewhat more useful than what I believe some IBM OS's do.
If you open a file, coding fixed instead of variable, or the other way
around, it will change the DSCB (file system file descriptor) but not
change the data.

-- glen

 

[Alan Balmer]

On the contrary, it specifically says that modes containing "b" open
binary files and modes without "b" open text files. It does not say
what happens if the named file is not of the correct type, so you get
undefined behavior if you use the return stream (if there is one).

Right. I hadn't noticed that (or it didn't register at the time),
being in the comments on the mode strings rather than in the function
description. I do think that the standard is unclear on this - given
the fuzziness of other relevant passages (including what Dan Pop
describes as a bug in the standard) and given the fact that other
behavior is explicitly identified as undefined, in the same paragraph.
Writing standards in English is non-trivial

(Late response - I took a day off from c.l.c to get some work done,
and came back to 400-odd messages, which is why I don't look at c.l.c
when I have a deadline <g>.)