Re: four changes of suspicious code (connectivity, sc, sw, editeng)

On Fri, May 08, 2015 at 01:01:40PM +0200, Lionel Elie Mamane wrote:
> On Fri, May 08, 2015 at 10:23:37AM +0200, Stephan Bergmann wrote:

> > One is in connectivity (Lionel, do you happen to know how to trigger that
> > code?):

> I uploaded a trigger case to
> http://people.freedesktop.org/~lmamane/trigger-java_io_Reader::available.odb

> Enable macros and run the macro "Main" in "Module1" in that file.

I tried it, and it doesn't work. In my tests, HSQLDB's
Clob.getCharacterStream().ready() returns true even at end-of-file.

So we might return "1" although we should return "0". Since the
documentation says: "Note: This method offers no information on
whether the EOF has been reached.", I suggest to just let it slide and
be like that: we return 1 even at EOF.

> This whole class is a mess. It confuses the notion of byte and the
> notion of character. In Java, and also in LibreOffice, a character is
> *two* bytes.

> I'm hesitating between:

> 1) "Unimplementing" getCharacterStream in the SDBC<->JDBC driver

> 2) Having it just call getBinaryStream (which uses java_io_InputStream)

> 3) Fixing java_io_Reader to properly handle the bytes <-> character
>    conversion, (which is annoying because it will have to do its own
>    buffering for when an odd number of bytes is read...).

I took option 3. Pushing soon.

-- 
Lionel