>> Does anybody know what the bug is in GDB that this works around? The >> workaround was added as part of r103439[1], which added support for >> multiple compilation units in one module. Does the bug ever affect >> modules with only a single compilation unit? >> > > It was not added by r103439. In fact, it was added by r32035. > > http://llvm.org/viewvc/llvm-project?view=rev&revision=32035 > > Unfortunately the log message and comments do not provide any hints, and I do not have access to the author anymore.Sounds like we should rip it out and see if there is any regression on the gdb testsuite. -Chris -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110413/35cce780/attachment.html>
On Apr 13, 2011, at 11:15 AM, Chris Lattner wrote:>>> Does anybody know what the bug is in GDB that this works around? The >>> workaround was added as part of r103439[1], which added support for >>> multiple compilation units in one module. Does the bug ever affect >>> modules with only a single compilation unit? >>> >> >> It was not added by r103439. In fact, it was added by r32035. >> >> http://llvm.org/viewvc/llvm-project?view=rev&revision=32035 >> >> Unfortunately the log message and comments do not provide any hints, and I do not have access to the author anymore. > > Sounds like we should rip it out and see if there is any regression on the gdb testsuite.I vaguely recall an issue (maybe a linker related) when the DWARF debug_line table was empty (source file with no functions). The fix was to pad out the line table with a couple of no-ops. -Nick -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110413/5e540415/attachment.html>
On 13 April 2011 20:02, Nick Kledzik <kledzik at apple.com> wrote:> I vaguely recall an issue (maybe a linker related) when the DWARF debug_line > table was empty (source file with no functions). The fix was to pad out the > line table with a couple of no-ops.Still, it should be a special case (or a better fix), The way it is looks strange. cheers, --renato
On Apr 13, 2011, at 11:15 AM, Chris Lattner wrote:>>> Does anybody know what the bug is in GDB that this works around? The >>> workaround was added as part of r103439[1], which added support for >>> multiple compilation units in one module. Does the bug ever affect >>> modules with only a single compilation unit? >>> >> >> It was not added by r103439. In fact, it was added by r32035. >> >> http://llvm.org/viewvc/llvm-project?view=rev&revision=32035 >> >> Unfortunately the log message and comments do not provide any hints, and I do not have access to the author anymore. > > Sounds like we should rip it out and see if there is any regression on the gdb testsuite.Agreed. r129461. - Devang -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110413/a9782abb/attachment.html>
On 13 Apr 2011, at 21:02, Nick Kledzik wrote:> I vaguely recall an issue (maybe a linker related) when the DWARF > debug_line table was empty (source file with no functions). The fix > was to pad out the line table with a couple of no-ops.Indeed, that was a linker bug rather than a gdb bug. See e.g. rdar:// problem/5611555 Jonas