https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=273956
--- Comment #8 from Paul Floyd <pjfl...@wanadoo.fr> --- (In reply to Ed Maste from comment #6) I know a thing or two about this. Initially Valgrind would only trigger reading debuginfo when it considered a file to have finished mapping. That used to mean seeing 1 RX and 1 RW section. Then lld 9 started writing 2 RW sections. Gory details here: https://bugs.kde.org/show_bug.cgi?id=452802 So now Valgrind takes a peak at the binary to count the RW PT_LOADs and then triggers the debuginfo when it has seen 1 RW and the right number of RW PT_LOADs. Pretty much all binaries on FreeBSD now have 1 RC and 2 RW. For instance readelf -l /bin/sleep RO 02 .interp .note.tag .dynsym .gnu.version .gnu.version_r .gnu.hash .hash .dynstr .rela.dyn .rela.plt .rodata .eh_frame_hdr .eh_frame RX 03 .text .init .fini .plt RW1 04 .ctors .dtors .jcr .init_array .dynamic .got RW2 05 .data .got.plt .bss (I added the 1st column) The valgrind tools are totally static so they son't have any segments like got plt readelf -l .in_place/memcheck-amd64-freebsd RO 01 .note.tag .rodata .eh_frame_hdr .eh_frame RX 02 .text RW1 03 .data .bss -- You are receiving this mail because: You are the assignee for the bug.
