the failure rate is high enough to be motivating. :-) i now have go 1.9 working for production builds. i will report back with results as soon as i have them.
- erik On Monday, December 4, 2017 at 6:16:29 PM UTC-8, Ian Lance Taylor wrote: > > On Sun, Dec 3, 2017 at 6:42 PM, <quan...@gmail.com <javascript:>> wrote: > > > > i'm running go 1.8.3 on linux. about 1 in ~10 billion calls (spread > across > > many machines), i get > > a backtrace that looks like the following: > > > > panic: runtime error: invalid memory address or nil pointer dereference > > [signal SIGSEGV: segmentation violation code=0x1 addr=0x28 pc=0x4c3406] > > goroutine 441026 [running]: > > bufio.(*Reader).ReadSlice(0x0, 0xa, 0x30, 0xc420256ec8, 0xc4201b9ef0, > > 0xc420315580, 0x0) > > #011/usr/lib/golang/src/bufio/bufio.go:316 +0x26 > > bufio.(*Reader).ReadLine(0x0, 0xa1f378, 0x30, 0xc4201b9ef0, 0x30, > > 0xc4201b9ef0, 0x30) > > #011/usr/lib/golang/src/bufio/bufio.go:367 +0x37 > > fakexpkg.(*Xpkg).tableParse(0xffffffffffffffff, 0x0, 0x0) > > <----- HERE > > #011/builddir/build/BUILD/posthoc-1.1/src/fakexpkg/xpkg.go:175 +0x86 > > created by fakexpkg.(*Xpkg).List > > #011/builddir/build/BUILD/posthoc-1.1/src/fakexpkg/xpkg.go:225 +0x2ac > > > > the code calling tableparse looks something like this. no references to > > anything > > table at the end are kept in other places. > > > > ret := make(chan []*Info, 1) > > go x.tableParse(bufio.NewReader(outpipe), ret) > > table := <-ret > > > > other c programs that i run on these same machines do not core dump at > all. > > > > since there is no use of the unsafe package anywhere in this program, > i'm > > confused as to > > how the Xpkg receiver could be -1 unless something has gone wrong with > the > > runtime. > > i feel like i must be missing something though, since it's never the > layer > > below. > > > > does anyone have any idea what's going on here, or some hints on > debugging > > this? > > Assuming that the race detector doesn't report any problems, this is a > strange example of memory corruption: not only is the receiver pointer > invalid, the two arguments to the function are nil even though the > code fragment shows that that is not possible. From your description > the problem is very very rare. How much time and energy do you have > for experimentation? If you have some, the first step is certainly to > try using Go 1.9, as various bugs have been fixed. If it still > happens the same way, the next step is to try to reduce the program to > a self-contained example that you can share. At a guess given that > three words appear to be corrupt, it may have something to do with the > way that goroutine arguments are saved. I don't see any way that > could fail, but then this is a very rare problem. > > Ian > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
