DamjanJovanovic opened a new pull request, #164: URL: https://github.com/apache/openoffice/pull/164
Our EditEngine, in main/editeng, has a number of containers for paragraphs, paragraph portions, lines, etc. These are based on svl's "PTRARR" type classes, which is just an array of some caller-defined class that grows as needed, like ::std::vector but hardcoded to 16 bit limits. Worse, a lot of code, even outside of main/editeng, is written under these assumptions, and passes 16 bit types and constants to EditEngine classes and retrieves 16 bit results. One particularly notorious bug emerging from this design, is that when you try to paste an HTML table of more than 65534 cells into Calc, the first 65534 cells are pasted, but all further cells are silently lost, something reported in at least 3 bugs in the last 18 years ([57176](https://bz.apache.org/ooo/show_bug.cgi?id=57176), [110486](https://bz.apache.org/ooo/show_bug.cgi?id=110486) and [117225](https://bz.apache.org/ooo/show_bug.cgi?id=117225)). This apparently happens because our HTML import is similar to XML DOM parsing, done in 2 phases, first parsing the results into memory, then processing those in-memory results to populate data into the spreadsheet (ScHTMLImport::WriteToDocument() in main/sc/source/filter/html/htmlimp.cxx). When parsing, each HTML cell (<TD>) becomes a new EditEngine paragraph, and after 65534 paragraphs, all further cells are ignored, so when it's time to populate the spreadsheet, the data isn't there to add. This enormous patchset changes the container for paragraphs to a new BaseList<T> class, which wraps ::std::vector and uses 32 bit integers to access it. All EditEngine methods are changed to take 32 bit paragraph index parameters, all EditEngine classes store any paragraph indices as 32 bit fields, all 16 bit paragraph index constants like 0xffff are changed to 32 bit 0xffffffff (and to a more readable constant like EE_PARA_NOT_FOUND), and all (known) calling code everywhere in OpenOffice is updated to take these changes into account. The bug is definitely fixed, all sample documents from Bugzilla paste fully and correctly. What is harder to prove is that nothing else broke. Much had to change, 136 files in 9 modules. OpenGrok helped tremendously in finding obscure places where EditEngine methods were getting called from. I've really tried to avoid introducing any new bugs, but it is hard to be sure with a change of this size. I am making a PR instead of directly pushing to allow you to test a lot ;). -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: dev-unsubscr...@openoffice.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@openoffice.apache.org For additional commands, e-mail: dev-h...@openoffice.apache.org
