0000608: All input/output as if by fgetc()/fputc() is wrong for wide streams

Austin Group Defect Tracker

Aardvark Mark IV

Viewing Issue Simple Details [ Jump to Notes ]

[ Issue History ] [ Print ]

Category

Severity

Type

Date Submitted

Last Update

0000608

[1003.1(2008)/Issue 7] System Interfaces

Objection

Error

2012-09-20 11:28

2019-06-10 08:55

Reporter

geoffclare

View Status

public

Assigned To

ajosey

Priority

normal

Resolution

Accepted

Status

Closed

Name

Geoff Clare

Organization

The Open Group

User Reference

Section

2.5

Page Number

490

Line Number

16788

Interp Status

---

Final Accepted Text

Summary

0000608: All input/output as if by fgetc()/fputc() is wrong for wide streams

Description

XSH section 2.5 says:

    "All input takes place as if bytes were read by successive calls
    to fgetc(); all output takes place as if bytes were written by
    successive calls to fputc()."

This is incorrect for wide-oriented streams, since 2.5.2 says that
byte input/output functions cannot be applied to a wide-oriented
stream.

The equivalent text in C99 is (7.9.13 items 11 and 12):

    "The wide character input functions read multibyte characters from
    the stream and convert them to wide characters as if they were
    read by successive calls to the fgetwc function. Each conversion
    occurs as if by a call to the mbrtowc function, with the
    conversion state described by the stream's own mbstate_t object.
    The byte input functions read characters from the stream as if by
    successive calls to the fgetc function."

    "The wide character output functions convert wide characters to
    multibyte characters and write them to the stream as if they were
    written by successive calls to the fputwc function. Each
    conversion occurs as if by a call to the wcrtomb function, with
    the conversion state described by the stream's own mbstate_t
    object. The byte output functions write characters to the stream
    as if by successive calls to the fputc function."

We can use this text almost as-is (needs "shallification" and changing
"multibyte character" to "character"), although it would also benefit
from the addition of forward references to 2.5.2.

This covers the byte input/output functions and the wide character
input/output functions, but there is also a problem with other
functions. For perror(), psiginfo() and psignal() there is a conflict
because "as if by fputc()" implies that the functions are required to
set an unoriented stream to byte orientation, but the descriptions of
the functions say that they do not change the orientation. For the
remaining functions getopt() and wordexp() there is the question of
whether the requirement to set an unoriented stream to byte
orientation matches existing practice. I have tested getopt() on
Solaris, HP-UX and Linux; on Linux it does set stderr to byte-oriented
when it writes an error message, but on the others it leaves stderr
unoriented. I did not test wordexp() since it is already well known
that error messages from wordexp() are often written by a child
process and therefore would not affect the orientation of stderr in
the calling process when that is the case.

A related issue is that if getopt() uses a byte output function
internally, and since these functions cannot be used on wide oriented
streams, applications need to ensure that getopt() will never try to
write a message to stderr if it has wide orientation. Similarly for
wordexp().

Desired Action

Replace the sentence:

    All input takes place as if bytes were read by successive calls
    to fgetc(); all output takes place as if bytes were written by
    successive calls to fputc().

with the following four paragraphs:

    The wide character input functions shall read characters from
    the stream and convert them to wide characters as if they were
    read by successive calls to the fgetwc() function. Each conversion
    shall occur as if by a call to the mbrtowc() function, with the
    conversion state described by the stream's own mbstate_t object
    (see [xref to 2.5.2]). The byte input functions shall read
    characters from the stream as if by successive calls to the
    fgetc() function.

    The wide character output functions shall convert wide characters
    to characters and write them to the stream as if they were
    written by successive calls to the fputwc() function. Each
    conversion shall occur as if by a call to the wcrtomb() function,
    with the conversion state described by the stream's own mbstate_t
    object (see [xref to 2.5.2]). The byte output functions shall
    write characters to the stream as if by successive calls to the
    fputc() function.

    The perror(), psiginfo() and psignal() functions shall behave as
    described above for the byte output functions if the stream is
    already byte-oriented, and shall behave as described above for the
    wide character output functions if the stream is already
    wide-oriented. If the stream has no orientation, they shall
    behave as described for the byte output functions except that
    they shall not change the orientation of the stream.

    Functions other than perror(), psiginfo() and psignal() that write
    to streams but are neither wide character output nor byte output
    functions (getopt() and wordexp()), shall behave as described
    above for the byte output functions, except that if the stream has
    no orientation, it is unspecified whether they set the stream to
    byte orientation or leave it with no orientation.

At page 1040 line 34848 section getopt() change:

    getopt() shall also print a diagnostic message to stderr in the
    format specified for the getopts utility.

to:

    getopt() shall also print a diagnostic message to stderr in the
    format specified for the getopts utility, unless the stderr stream
    has wide orientation in which case the behavior is undefined.

After page 1043 line 34980 section getopt() add a new paragraph
to APPLICATION USAGE:

    Applications which use wide character output functions with stderr
    should ensure that any calls to getopt() do not write to stderr,
    either by setting opterr to 0 or by ensuring the first character
    of optstring is always a <colon>.

At page 2258 line 71006 section wordexp() change:

    wordexp() may write messages to stderr if syntax errors are
    detected while expanding words.

to:

    wordexp() may write messages to stderr if syntax errors are
    detected while expanding words, unless the stderr stream has
    wide orientation in which case the behavior is undefined.

After page 2259 line 71038 section wordexp() add a new paragraph
to APPLICATION USAGE:

    Applications which use wide character output functions with stderr
    should ensure that any calls to wordexp() do not write to stderr,
    by avoiding use of the WRDE_SHOWERR flag.

Notes
(0001370) geoffclare (manager) 2012-09-20 13:31	Another thought on wordexp()... If the error message is written by a child process, presumably it would always be written to fd 2, which in rare circumstances might not be the same as fileno(stderr). Or do wordexp() implementations check for this and when necessary add "2>&d" (where d is the value of fileno(stderr)) to the shell command they execute?

(0001382) eblake (manager) 2012-09-26 16:18	0000555 addresses the concerns about wordexp() using fd 2 - applications are already required to take care that fileno(stderr) is always 2 if they want conforming behavior.

Issue History
Date Modified	Username	Field	Change
2012-09-20 11:28	geoffclare	New Issue
2012-09-20 11:28	geoffclare	Status	New => Under Review
2012-09-20 11:28	geoffclare	Assigned To	=> ajosey
2012-09-20 11:28	geoffclare	Name	=> Geoff Clare
2012-09-20 11:28	geoffclare	Organization	=> The Open Group
2012-09-20 11:28	geoffclare	Section	=> 2.5
2012-09-20 11:28	geoffclare	Page Number	=> 490
2012-09-20 11:28	geoffclare	Line Number	=> 16788
2012-09-20 11:28	geoffclare	Interp Status	=> ---
2012-09-20 13:31	geoffclare	Note Added: 0001370
2012-09-26 16:18	eblake	Note Added: 0001382
2012-09-26 16:23	Don Cragun	Status	Under Review => Resolved
2012-09-26 16:23	Don Cragun	Resolution	Open => Accepted
2012-09-26 16:23	Don Cragun	Desired Action Updated
2012-09-26 16:24	Don Cragun	Tag Attached: tc2-2008
2019-06-10 08:55	agadmin	Status	Resolved => Closed

Relationships