From esr at thyrsus.com Fri Nov 7 19:32:39 2003 Content-Type: multipart/mixed; boundary="===============1816682998778173254==" MIME-Version: 1.0 From: Eric S. Raymond To: devel at lists.fedoraproject.org Subject: Re: RPM submission script Date: Fri, 07 Nov 2003 19:38:09 -0500 Message-ID: <20031108003809.GB17763@thyrsus.com> In-Reply-To: 1068249597.14565.8.camel@binkley --===============1816682998778173254== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable seth vidal : > > One of the fields handed to the CGI may be a URL pointing to an RPM. > > It seems to me that the encoding of strings *in the RPM* is not an issu= e at > > all for bug-bugzilla. Only the contents of the fields in the job card > > is an issue. I don't understand why the requirement that those be UTF-8 > > should ever be onerous. = > = > b/c the content you're providing may well be data that is in an rpm > header or in the output of rpm -qi somepkg. > = > as the example I gave to Nicholas, rpm -qi cups - read the description > field, note the (R), notice that's it is 1 character, not 3. > = > your program can force the conversion and replace the characters outside > the 128 range with ?, but in the case of filenames that are using > non-utf-8 encoding you WILL lose some data. Ahhhh. Now I see. This is not actually an issue sbout bug-bugzilla at all. It's an issue about how fedora-submit will transcode non-ASCII stuff that it mines out of RPM metadata fields. I was aware of this problem. fedora-submit will have to call iconv in certain circumstances to mung all the fields into UTF-8. Sometimes it will fail and fedora-submit will error out as a result. Life is hard. -- = Eric S. Raymond --===============1816682998778173254==--