Ash on my head, everyone stop scratching your heads: I found the real culprit and it's me, me, me and me alone.

Since the beginning of times and/or Trados Studio (or at least since the words "Fill your computers with software and subdue it" were uttered) I have a segmentation rule for tabs on all my TMs. It goes like this: no matter what comes before or after, a tab is a segment break.

It actually had crossed my mind this could be the crux of the matter and I even removed the rule for a test but obviously I must have made a mistake. New day, new chance to better ourselves I tried again with a freshly minted TM and all the default settings and, hey presto, the matter was solved. Special thanks once more to Kelly Edward who put me on the right path.

(Still curious though why the error occurs and why not during the conversion to xliff - where AFAIK also the segmentation happens - but while interpreting the xliff for analysis, pre-translation and what not)

Angela Sigee 

Hi Angela,

I'm a little confused at your last post, though I'm pleased you found a solution!

You thanked Kelly Edward for putting you on the right track... I'm pleased that he did but were you in private communication with each other?

If not, I am unsure how his last post gave you the solution.

If so, can you share here the full solution to your problem ... always useful for future users looking for answers.

You said 'again with a freshly minted TM' - how does that indicate that it's your segmentation rule that was at fault? Was it the only non-default rule?

Just asking in case there is more information that could be useful to others... and because I'm a pedant...

Glad your problems were sorted either way...

All the best,

Ali

Hi Alison,

No there wasn't any private communication. The mere fact that Kelly couldn't reproduce the error made me rethink.

The "freshly minted TM" is indeed one with the default settings, so no custom segmentation rule.
I also run a test with the added Tab segmentation rule (Anything-Tab-Anything) to confirm my suspicion that indeed it was that rule which caused the error.

I the meantime I also noticed that the segment markers (<mrk>) are missing from the xliff when the error occurs, which must be why - while the xliff may still be valid xml - it is not valid for analysis, pre-translation etc.

I would maintain though that it is a bug (it shouldn't happen), albeit most likely a minor one. On the other hand it also might point to a bigger problem. It looks like something quite hard to catch without a test case.

 Angela Sigee 

Hi Angela,

Thanks for your reply. Paul, is it likely that the issue described above is a bug?

All the best,

Ali

 Angela Sigee 

Hi Angela,

Thanks for your reply. Paul, is it likely that the issue described above is a bug?

All the best,

Ali