From mboxrd@z Thu Jan 1 00:00:00 1970 From: Phil Edwards To: law@redhat.com, bkoz@redhat.com Cc: Janis Johnson , Jakub Jelinek , Franz Sirl , Mark Mitchell , gcc@gcc.gnu.org, "Kevin B. Hendricks" Subject: Re: GCC 3.2 Date: Mon, 19 Aug 2002 11:41:00 -0000 Message-id: <20020819144122.B23804@disaster.basement.lan> References: <20020812164321.A4428@us.ibm.com> <200208131649.g7DGnkK09005@porcupine.slc.redhat.com> X-SW-Source: 2002-08/msg01133.html On Tue, Aug 13, 2002 at 10:49:46AM -0600, Jeff Law wrote: > Right. That's one of the reasons why I mentioned in an earlier message > that we need a test which verifies that all of the exported functions > in libstdc++ (and libgcc) are maintained over time -- which has the > side effect of being a reasonable tester for name mangling changes. > We need ways to verify structure layouts aren't changing, etc etc. In message < 20020812164321.A4428@us.ibm.com >, Janis Johnson writes: > >and new compilers available when running the tests. I'm not at all sure > >how to set up such tests for use with GCC. The test harness needs to > >know about two compilers under test rather than one, For some time now my nightly autocrasher[*] has built both the current trunk and the current 3.whatever-the-previous-release-was branch. I've started comparing exported symbol information from the libstdc++.so's from both builds. (See the libstdc++ list archives of the last week or so; posts from Benjamin and Ulrich and myself.) I've verified that changing the size of an exported symbol triggers the regression detection. Work yet to be done: 1) Test for layout changes, not just size changes. 2) Find a way of putting this into the v3 repository. I've no ideas on (1). For (2) I'll be posting some thoughts on the v3 list later. Phil [*] "autobuilder" is just too positive and uplifing a name for something as hacky as what I wrote. -- I would therefore like to posit that computing's central challenge, viz. "How not to make a mess of it," has /not/ been met. - Edsger Dijkstra, 1930-2002