|
From: Tony R. <ton...@bu...> - 2006-09-06 07:50:50
|
Le mardi 05 septembre 2006 =E0 07:19 +1000, Nicholas Nethercote a =E9crit : > On Mon, 4 Sep 2006, Tony Reix wrote: > > If I remember well, Helgrind provides its own thread library. So it > > helps to find some kinds of bugs, but it cannot help in many complex > > cases related to NPTL details. >=20 > That was true then, but Valgrind now doesn't replace the native thread=20 > library. Oh. So I have to refresh my knowledge about Helgrind and check that my claims are still correct. > It looks like an interesting tools, but I don't see why it should be part= of=20 > the Valgrind distribution. If it was built with Valgrind, maybe, but it=20 > seems to be unrelated apart from involving debugging. The 2 basic ideas of my email are 1) that NPTL does not provide the tools required by Distros in order to guarantee a good support of NPTL and multi-threading for Linux customers, and 2) that understanding where is the problem may be a deadful pain. At the very beginning of PTT, I was in contact with IBM JVM guys trying to understand the cause of a very annoying problem. These guys spent days and weeks to locate and understand the cause. Not all users of NPTL have this level of knowledge of Linux and NPTL. So, such a tool is very useful. Why add PTT to Valgrind ? Because it would be a good place to maintain it and to make it evolve, because many programmers/maintainers of multi-threading applications are surely looking at Helgrind when they encounter a problem. PTT is aimed not only to help programmers debug their own program, but also to help maintainers of complex multi-threading applications. Part of what is sometimes called RAS (Reliability, Availability, Serviceability). Tony |