Community technical support mailing list was retired 2010 and replaced with a professional technical support team. For assistance please contact: Pre-sales Technical support via email to sales@march-hare.com.
We have just moved to cvsnt from traditional cvs and have met a weird problem trying to change subst mode from -kb to -ku. Assume our unicode.xsl constains text in UTF-16 and stored at server as binary. 1. Change subst mode copy unicode.xsl unicode.bak cvs admin -ku unicode.xsl cvs up unicode.xsl File become double-sized and corrupted. Apparently, that's ok since server considers it is UTF-8. 2. Trying to commit correct file copy unicode.bak unicode.xsl cvs ci -m"correcting contents after admin -ku" unicode.xsl commit completes correct, telling us that new revision is created. But that revision is exactly as previous and file in sandbox becomes double-sized and corrupted again! cvs up -kb -P brings us original UTF-16 file, as expected. 3. We managed to correct file contents only by commiting original but slightly changed file. Though, it was not 100% remedy - sometimes (at some files) it required two such commits to make all right. NB: Trying to reproduce this situation on cvsnt server, I got it all ok. Apparently, it only happens when file was initially added at traditional cvs. So, the question is: are we do something wrong? If not, is there more straight way to correctly convert binary files to unicode? CVSNT ver 2.0.41a at server (Linux) and client (Win2K). Thanks in advance, Iakov.