Re: lib/47602 (getwc() modifies input instead of returning EILSEQ.)

[yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi)]

The following reply was made to PR lib/47602; it has been noted by GNATS.

From: yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi)
To: christos%zoulas.com@localhost
Cc: gnats-bugs%NetBSD.org@localhost, lib-bug-people%netbsd.org@localhost, 
netbsd-bugs%netbsd.org@localhost,
        gnats-admin%netbsd.org@localhost, yamt%NetBSD.org@localhost
Subject: Re: lib/47602 (getwc() modifies input instead of returning EILSEQ.)
Date: Tue, 21 Jan 2014 06:12:40 +0000 (UTC)

 > On Jan 21,  2:07am, yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi) 
 > wrote:
 > -- Subject: Re: lib/47602 (getwc() modifies input instead of returning EILSEQ
 > 
 > | do you mean to expect the "c == obuf[i]" test always true?
 > 
 > If that's not true, many things turn quickly very complicated, because
 > you cannot read and write a file using getwc() and write it using putwc()
 > and expect it to work. If I know that wchar_t is 2 bytes and I call 2
 > getc()'s I can expect to get the next two bytes from the file I think.
 
 it's expected to work if you do putwc and getwc within the same locale.
 
 > 
 > | while obuf[] is wchar_t for big5 locale (right?), what getwc returns
 > | is wchar_t for another locale.  comparing them with "==" doesn't make
 > | much sense.
 > 
 > Right, but I am not sure if this is such a great idea.
 > 
 > | (i'm not sure about the intention of the mixed use of getwc and getc
 > | in the program, though.)
 > 
 > I did not write this, but I think the logic is, read with getwc() and
 > if that fails, skip the next 2 bytes with getc() and keep going. How
 > else do you recover? Bail for the rest of the file, because two bytes
 > got corrupted?
 
 thanks for explanation.
 
 you can probably recover in that way except that it's in general
 difficult to make a correct guess how many bytes to skip.
 besides that, probably it's safer to use fgetpos/fsetpos to recover
 the position.
 
 C locale is not rich enough for programs like editors.
 
 YAMAMOTO Takashi
 
 > 
 > christos