[Libreoffice] [Bug 42865] MySQL native driver: free() on non-heap pointer; in debug mode abort()

Lionel Elie Mamane lionel at mamane.lu
Mon Nov 14 03:08:46 PST 2011


(We are having half the discussion on the ML and half of it in the bug...)

On Mon, Nov 14, 2011 at 11:44:47AM +0100, Lionel Elie Mamane wrote:
> On Mon, Nov 14, 2011 at 10:47:31AM +0100, Stephan Bergmann wrote:

> Now that you mention it, my *system* libmysqlcppconn.so also links
> against the symbol:

> $ objdump -T /usr/lib/libmysqlcppconn.so|grep _rep_st
> 0000000000000000      DO *UND*	0000000000000000  GLIBCXX_3.4 _ZNSs4_Rep20_S_empty_rep_storageE

> But the LibreOffice-*internal* libmysqlcppconn defines the symbol as
> "globally unique":

> $ objdump -T ~/src/libreoffice/core/mysqlcppconn/unxlngx6/lib/libmysqlcppconn.so|grep _rep_st
> 000000000031bca0 u    DO .bss	0000000000000020  Base        _ZNSs4_Rep20_S_empty_rep_storageE

> would that be a change introduced by -D_GLIBCXX_DEBUG?

Yes! If I recompile my distro package with -D_GLIBCXX_DEBUG, I get the
"u" symbol again:

$ # hack build system to add -D_GLIBCXX_DEBUG
$ debuild -uc -us -b
$ objdump -T ~/src/mysqlcppconn/mysql-connector-c++-1.1.0/debian/libmysqlcppconn5/usr/lib/libmysqlcppconn.so.5.1.1.0 | grep -i _rep_st
000000000035fd60 u    DO .bss	0000000000000020  Base        _ZNSs4_Rep20_S_empty_rep_storageE

-- 
Lionel


More information about the LibreOffice mailing list