Thanks for investigating! You seem to have done a very thorough job. The problem is replicable on another installation of ArchLinux, so I guess it does have to do with the build. I think I ll continue to use my broken patch until ArchLinux releases a new build of GCC. On Thu, Jul 2, 2009 at 6:50 AM, Joseph Prostko <joe.prostko+haiku@xxxxxxxxx<joe.prostko%2Bhaiku@xxxxxxxxx> > wrote: > On Wed, Jul 1, 2009 at 2:03 PM, Joseph > Prostko<joe.prostko+haiku@xxxxxxxxx <joe.prostko%2Bhaiku@xxxxxxxxx>> > wrote: > > On Wed, Jul 1, 2009 at 11:55 AM, Travis D. Reed<tdreed@xxxxxxxxx> wrote: > >> Here's the relevant compiler output. I'm using gcc (GCC) 4.4.0 20090526 > >> (prerelease), which is what ArchLinux requires at the > >> moment. > > > > > I'm currently testing other > > breakage, so will wait to update my Linux compiler to do further > > testing in this regard. > > > > Okay, first of all, the "other breakage" I mentioned above was human > error on my part, so I didn't have to investigate that too long. :) > > Back on topic though, I messed around with my host compiler a bit. I > first installed the latest available GCC 4.4.0 package for Ubuntu, but > encountered no problems with the Haiku build. After that, I built the > latest GCC 4.4 snapshot from the other day, and it built Haiku just > fine as well. I rebuilt the buildtools and all of Haiku each time > just to make sure the fs_shell objects were getting rebuilt from > scratch. Actually, I blew away all of my generated dir too throughout > this process, just to be extra sure. > > So anyways, I'm not sure what to think, as currently both my host and > cross-compiler are both at the latest GCC 4.4 snapshot level, and I > don't run into the problem you're having. I am using the latest trunk > with the latest 32bit-w_char branch changes merged in, but that > shouldn't be relevant in this case. Come to think of it, some time > ago I was building Haiku in Haiku with the native GCC 4.4.0 I was > testing, and I didn't have any hiccups with it bombing out either. It > doesn't appear as if any changes to fs_shell in this regard since then > would now cause an issue either. > > I also logged the build output to see if I noticed anything odd, but I > never encounter the problem with string.o like you get. It builds > cleanly here with no warnings or anything. I did get warnings for > "invalid access to non-static data member" several times in vfs.cpp, > but nothing that seems to be related to this problem. > > I have no more ideas at the time, really. I thought my last email was > on the right track, but perhaps not. Maybe it's just something > specific with the Arch Linux implementation or environment? > > Axel may have a better clue now if he dissects anything relevant from > my ramblings. :) > > - joe > > -- --Travis D. Reed