> On June 28, 2013, 2:53 p.m., David Faure wrote: > > Independently from this patch, I was thinking of a completely different > > approach to solve non-utf8 encodings in file:/// urls: using > > KRemoteEncoding in kio_file, just like we do in other slaves like kio_ftp. > > I.e. letting the user select the encoding by hand for the current > > directory. This way, not only would files be usable, but they could even > > appear correctly. However it requires manual user intervention, but that's > > the case anyway for FTP dirs that use another locale etc. > > On the other hand it moves the fix "up to KIO", so it wouldn't work for > > other QFile uses.. [hmm, or anywhere where KIO has a fast path for local > > files to avoid calling kio_file.... Ouch, that might kill this idea > > completely, in fact]. Well, food for thought, then.
Using anything but the exact byte sequence in URLs breaks http://freedesktop.org/wiki/Specifications/file-uri-spec/. That spec and RFC 3987 put together limit KDE to exactly one local filesystem encoding: UTF-8. - Thiago ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://git.reviewboard.kde.org/r/110043/#review35232 ----------------------------------------------------------- On June 28, 2013, 9:28 a.m., Róbert Szókovács wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://git.reviewboard.kde.org/r/110043/ > ----------------------------------------------------------- > > (Updated June 28, 2013, 9:28 a.m.) > > > Review request for kdelibs and Thiago Macieira. > > > Description > ------- > > This patch works around the problem of filenames that are not valid UTF8 > strings: in KLocalePrivate::initFileNameEncoding() KDE sets the QFile's > encoding/decoding function, to to/fromUTF8() in QString, which in turn calls > QUtf8's converter function (QUtf8 is not exported to developers, so I had to > use an inefficient method, I think it would be better if we could use the > state parameter for error detection). I replaced this with the said > functions' copy/pasted version and changed it, so when it encounters an > invalid UTF8 string, it will encode it byte by byte, mapping the lower 128 > their normal unicode place and the upper 128 to U+18000-U+1807F, and of > course the decoder reverses it. > To make this actually work you have to define the KDE_UTF8_FILENAMES > enviroment variable to a specific value ("broken_names"). > > To test it, do the following: .kde/env/KDE_UTF8_FILENAMES.sh with this > content: > export KDE_UTF8_FILENAMES=broken_names > logout, login, try dolphin on faulty files. (instead of the usual boxed "?" > you'll see just boxes) > > > This addresses bug 165044. > http://bugs.kde.org/show_bug.cgi?id=165044 > > > Diffs > ----- > > kdecore/localization/klocale_kde.cpp b010e74 > > Diff: http://git.reviewboard.kde.org/r/110043/diff/ > > > Testing > ------- > > > Thanks, > > Róbert Szókovács > >
