Nikolay Sivov nsivov@codeweavers.com wrote:
On 3/24/21 10:42 AM, Dmitry Timoshkov wrote:
What's the status of the first two patches in the series?
Encoding you used seems to be a reason for broken Vista case. Since this does not test anything encoding-specific, utf-8 will make it work without broken() cases. Test itself belongs to test_save().
Which encoding would you suggest to use for the test then?
Regarding 2/3, the reason to unlink originally was to fix duplicated prologs, occasionally appearing in output. I don't remember details now, it's possible it was about first loading document, and then inserting prolog PI.
I guess since the patch in 2/3 doesn't break existing tests then the problem you mentioned might be fixed already, so it might be safe to assume that the patch is correct.
This is again a case where implementing own dumping logic will help, both with prolog and newlines.
Then again, the patch could be a light alternative to a full rewrite, especially if nobody is planning to work on this in near future.