[Libreoffice-bugs] [Bug 137737] REPLACE not working if there is a footnote or endnote anchor at end of paragraph

[bugzilla-daemon at bugs.documentfoundation.org]

https://bugs.documentfoundation.org/show_bug.cgi?id=137737

--- Comment #17 from Mike Kaganski <mikekaganski at hotmail.com> ---
DoSearch in sw/source/core/crsr/findtxt.cxx uses lcl_CleanStr, which does some
utter magic with special characters in the text. Namely, in case of
RES_TXTATR_FTN (among other things), it replaces it with '\x7f' in the middle
of the text, and drops in the end. The input looks like

  "He heard quiet steps behind him. That didn't bode well. Who could be
following\x01 him this late at night and in this deadbeat part of town."\x01

and the output looks like

  "He heard quiet steps behind him. That didn't bode well. Who could be
following\x7f him this late at night and in this deadbeat part of town."

That latter string arrives to the TextSearch::searchForward's searchStr; and in
it, the "$ matches (because the footnote anchor is dropped). Note how you will
be unable to find regex 'following ' (without quotes; i.e. the word "following"
followed by a space) in this text - because there is a footnote anchor between
the word and the space.

Personally I would simply not strip the anchor from the string; but that needs
a way to treat those meta-characters specially - having some syntax for them;
another option would be to remove them *everywhere*, not only at the end (so
that the 'following ' example would work). But I digress.

OTOH, ReplaceBackReferences (again in sw/source/core/crsr/findtxt.cxx) does not
pre-process the string obtained from the paragraph using that lcl_CleanStr, and
passes the original string to TextSearch::searchForward as is. Hence the
difference between the results of find and replace. The replace previously
relied on the "find" step limiting the endPos, and so the truncation happened
inside TextSearch::searchForward using that position; and now the function sees
the actual anchor following the position, and the regex does not match.

Trying to find a way to "synchronize" this - likely the "natural" option is to
make ReplaceBackReferences use lcl_CleanStr.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice-bugs/attachments/20210826/8b8661c0/attachment.htm>