tree 94f15675b30aafa68b84670c581cb7c5de4cb3a6
parent 7388bd02f160a35d06b58e57f6374780fe4bafd3
author Nico Weber <thakis@chromium.org> 1701469500 +0000
committer Pdfium LUCI CQ <pdfium-scoped@luci-project-accounts.iam.gserviceaccount.com> 1701469500 +0000

Make WideString::ToUTF16LE() do surrogate splitting if needed

Should make text strings with characters outside of the BMP show up
correctly on non-Windows platforms.

(Because e.g. PDFiumEngine::TraverseBookmarks() calls
FPDFBookmark_GetTitle() expecting to get a std::16string, and
FPDFBookmark_GetTitle() calls Utf16EncodeMaybeCopyAndReturnLength()
which calls WideString::ToUTF16LE() which currently chops off the
two upper bytes of wchar_t on platforms where wchar_t is 4 bytes --
i.e. on non-Windows.)

It's still a bit unfortunate that:
* ToUTF16LE() is a method on WideString, since that means we can't
  use it when we only have a WideStringView, such as in PDF_EncodeText()
  (which wants UTF16BE, but we could call UTF16LE and swap bytes).
  We might want to move the method to be a static thing on ByteString
  instead: ByteString::UTF16LEFromWide(WideStringView)
* ToUTF16LE() adds a zero-terminator, making it nonsymmetrical with
  FromUTF16LE. Some places (AsFPDFWideString()) want to convert the
  result to a FPDF_WIDESTRING which is nul-terminated. Maybe there
  should be an arg for opting in to the zero termination.

But again, that's for separate CLs.

Bug: pdfium:2105
Change-Id: I7aa1f1aa96f7b490c0c4435577fffe82a91809d9
Reviewed-on: https://pdfium-review.googlesource.com/c/pdfium/+/114130
Commit-Queue: Tom Sepez <tsepez@chromium.org>
Auto-Submit: Nico Weber <thakis@chromium.org>
Commit-Queue: Nico Weber <thakis@chromium.org>
Reviewed-by: Tom Sepez <tsepez@chromium.org>