Bug Description:
If a TM import file contains 2-byte UTF-8 sequence at an offset multiple of 0x4000 followed by another multibyte UTF-8 sequence, then the character encoded by the second sequence is corrupted after the import.
Repro Steps:
1) Create a new TM with the source and target languages set to "English (United States)".
2) Import the file test.txt with default import options. The last line of the preamble in this file is padded with spaces so that the only translation unit in the file crossed the 0x4000 offset.
3) Search for the "test" in TM.
Result: The third character in the translation is corrupted.
Impact:
If exported software strings are localized in Trados, this peculiar bug will make into the software. The bug unexpectedly appears on clean files, and therefore hard to discover and eliminate.
Software configuration: Windows Vista, SDL Trados Translator's Workbench 8.3.0.863. The error does not occur on Windows XP.
Workaround: Please, contact Logrus engineers for workaround on this.