Subject: Re: Text import broken
From: Thomas Fletcher (thomasf@qnx.com)
Date: Fri Apr 20 2001 - 07:26:39 CDT
On Fri, 20 Apr 2001, =?iso-8859-1?q?Andrew=20Dunbar?= wrote:
> Sometime over the past 2 to 4 days text import has
> been broken at least on Windows builds.
>
> I have looked into the problem and I've found that
> when the 8 bit characters are converted to the 16-bit
> internal representation that the two bytes are swapped
> incorrectly.
>
> Thus a space character is converted from
> 0x20 to 0x2000 instead of 0x0020.
>
> Other importers are not affected. Text export may
> have
> the same problem.
>
> Not sure about other platforms. Somebody please
> check.
Dom and I had a little crash session with this last night.
The Unix builds seem to not be affected .. my suspicion
is that it is because they are using the system libiconv
while QNX (who is broken) and Windows are using the tree
version of libiconv. Now this didn't _used_ to be broken
(ie two weeks ago before I went away for a week) so this
morning I'm rolling back to April 6th and re-building
to check and see if the problem is present in that build.
The problem seems to be that iconv() isn't generating the
right wc->mb translation in the mbtowc. According to
dom's unix code our Encoding is the same so I'm hoping
to rule out any recent changes in the Encoding manager.
The short of it is that I'm a bit dumbfounded at the moment
and backtracking to see how this happened. Whatever I
find I'll report and should be able to address the windows
problems as well (BeOS too I suspect).
Thomas
-------------------------------------------------------------
Thomas (toe-mah) Fletcher QNX Software Systems
thomasf@qnx.com Neutrino Development Group
(613)-591-0931 http://www.qnx.com/~thomasf
This archive was generated by hypermail 2b25 : Fri Apr 20 2001 - 07:26:37 CDT