Invalid byte 2 of 3-byte UTF-8 sequence.

Hey Stefan, please provide more info *g*. Interesting would be (as you of course know) filename, description etc. of the changeset :-).

This seems to be due to the locale stuff that is especially painful when using XML parsers. I've already thought about writing a custom parser that just takes the output and doesn't bother about locales...

The only solution to this that I can think of is to create a dialog to choose an encoding per hg root, e.g. UTF-8 or ISO-8859-15 that can then be used as preamble for the XML document format. Everything else won't work, as we don't know whether somebody used an e.g. cyrillic codeset for creating a changeset description. And Mercurial will only save the description as byte array, so no help from that side...

Any news on this. Does it still occur? I've submitted a changeset a few weeks ago, that might have fixed it...

Replying to bastiand:

Any news on this. Does it still occur? I've submitted a changeset a few weeks ago, that might have fixed it...

I just did a fresh install of Mercurial (1.0.2) and Mercurial Eclipse (1.1.867) on Windows XP and I am seeing this when doing a pull. Note that the pull seems to succeed nonetheless. I do not see it when trying to show history.

I do NOT see this on Linux - Mercurial 1.0.1 and Mercurial Eclipse (1.1.867).

The fix is only available in the current beta so far... But how does the changeset xml looks like that makes Mercurial under windows cry?

I'm using Mercurial 1.0.2 and MercurialEclipse? 1.1.19 and the bug is still there... :-/

com.sun.org.apache.xerces.internal.impl.io.MalformedByteSequenceException: Invalid byte 2 of 3-byte UTF-8 sequence.
at com.sun.org.apache.xerces.internal.impl.io.UTF8Reader.invalidByte(UTF8Reader.java:674)
at com.sun.org.apache.xerces.internal.impl.io.UTF8Reader.read(UTF8Reader.java:398)
at com.sun.org.apache.xerces.internal.impl.XMLEntityScanner.load(XMLEntityScanner.java:1742)
at com.sun.org.apache.xerces.internal.impl.XMLEntityScanner.peekChar(XMLEntityScanner.java:487)
at com.sun.org.apache.xerces.internal.impl.XMLDocumentFragmentScannerImpl$FragmentContentDriver.next(XMLDocumentFragmentScannerImpl.java:2679)
at com.sun.org.apache.xerces.internal.impl.XMLDocumentScannerImpl.next(XMLDocumentScannerImpl.java:648)
at com.sun.org.apache.xerces.internal.impl.XMLNSDocumentScannerImpl.next(XMLNSDocumentScannerImpl.java:140)
at com.sun.org.apache.xerces.internal.impl.XMLDocumentFragmentScannerImpl.scanDocument(XMLDocumentFragmentScannerImpl.java:510)
at com.sun.org.apache.xerces.internal.parsers.XML11Configuration.parse(XML11Configuration.java:807)
at com.sun.org.apache.xerces.internal.parsers.XML11Configuration.parse(XML11Configuration.java:737)
at com.sun.org.apache.xerces.internal.parsers.XMLParser.parse(XMLParser.java:107)
at com.sun.org.apache.xerces.internal.parsers.AbstractSAXParser.parse(AbstractSAXParser.java:1205)
at com.vectrace.MercurialEclipse.commands.AbstractParseChangesetClient.createMercurialRevisions(AbstractParseChangesetClient.java:447)
at com.vectrace.MercurialEclipse.commands.HgLogClient.getProjectLog(HgLogClient.java:129)
at com.vectrace.MercurialEclipse.history.MercurialHistory.refresh(MercurialHistory.java:179)
at com.vectrace.MercurialEclipse.history.MercurialHistoryPage$RefreshMercurialHistory.run(MercurialHistoryPage.java:107)
at org.eclipse.core.internal.jobs.Worker.run(Worker.java:55)

Seems to be a problem with non-ASCII charsets. Google says to problem occurs when you try to parse a UTF-8 document with ISO-8859-1 characters. I don't really understand the problem. When my commit message in Mercurial contains non-ASCII chars shouldn't they be UTF-8 encoded?

This patch solve the problem for me. Mercurial uses UTF-8 as default and so do all modern Linux distributions. But Windows XP doesn't...

Uh...the patch only contains an import. I can't imagine that an import solves the problem. I still haven't been able to reproduce. Could you please paste a changeset summary that causes this bug to trigger, together with local codepage and repository codepage you used? I assume this can maybe be fixed by using the HGENCODING variable.

Other patch contained garbage

Commit XML from the debugger

The commandline says CP437, but the system has probably CP1252 (the default, I didn't change anything). I didn't change anything on the Mercurial settings too. So I assume that Mercurial uses the default (UTF-8).

what's the command line output?

In Eclipse? The Changeset XML in the Mercurial Log and the Exception in the Eclipse Error Log...
I don't think the HGENCODING will help here, because that changes the encoding of Mercurial and not the system. It would only help if Mercurial converts his internal UTF-8 changesets into changesets which use the system encoding for the system and vice versa (the Repositories I push and pull need UTF-8 changesets)...
The patch I submitted should definitely solves the problem. I just tested it under another system...

Ops, forgot to log me in. BTW: How can I add me to the CC list?

Yeah, the login issue... I always forget it myself as well... You can either create an Assembla account and log yourself in with it or add your e-mail as CC. That should normally work. If you got an Assembla account, you'll receive notifications when somebody adds to a ticket that you've responded to.

I'll test your page as soon as I got Mercurial working on my computer. Seems a bit more complicated to run current crew on MacOS compared to Ubuntu...

Btw., page = patch.

A slightly modified form of your patch is now in mainline. Thanks for providing it! Btw., the patchfile failed to apply on my repository - did you use an old MercurialEclipse? version as base?