On 14/5/19 9:41 am, Joel Sherrill wrote: > On Mon, May 13, 2019, 6:04 PM Chris Johns <chr...@rtems.org > <mailto:chr...@rtems.org>> wrote: > > Hi, > > I chased down a buffer overwrite bug in libdl (#3746). On the powerpc and > sparc > the test dl05.exe crashed and on xilinx_zynq_a9_qemu I got a heap > protection > error on free. This was really nice however it is a mission to take that > error > message and figure out the block address that is the cause. A had to add a > couple of additional prints to get the block address and with that I > could set a > watch point that triggered on the invalid write. The patch to fix the bug > was > the easy part. > > Would a bad block variable set by the error handler that can be inspected > by a > debugger or exit handler be useful? > > I would like to see the ability to register a handler for various heap issues. > And then pass all the info you have to it.
Oh yes that is a great idea. Chris _______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel
