|
From: Philip S. <ph...@st...> - 2010-01-21 09:47:22
|
Hello, I was wondering how I can achieve an immediate abort at the first valgrind error v.s. waiting for the process to complete and then grepping the error log. This is important in automatic test case simplification where the error must be reported immediately so that the simplification algorithm can move forward. Any pointers on where in Valgrind's code it would be best to add an extra exit(), figuratively speaking, would be appreciated. Thank you! Philip Stoev |
|
From: Ashley P. <as...@pi...> - 2010-01-21 10:06:38
|
On 21 Jan 2010, at 09:47, Philip Stoev wrote: > This is important in automatic test case simplification where the error must > be reported immediately so that the simplification algorithm can move > forward. > > Any pointers on where in Valgrind's code it would be best to add an extra > exit(), figuratively speaking, would be appreciated. At a guess I'd look for the attach_gdb_on_error code and work from there. As it's possible to specify your own gdb command it might be possible to do this without modifying valgrind by writing a script which kills it's parent process and specifying this script as the "gdb command" to valgrind, you'd be on your own with this but it might be easier to maintain than a valgrind patch. Ashley, |