Hi Zubin, On Sat, Aug 09, 2014 at 09:46:42AM +0530, Zubin Mithra wrote: > On Sat, Aug 9, 2014 at 5:05 AM, Dmitry V. Levin <[email protected]> wrote: > > On Tue, Jul 01, 2014 at 03:13:21PM +0530, [email protected] wrote: > > > * defs.h (show_fd_path): Change type to unsigned int. > > > * strace.c (show_fd_path): Update usage to count y flag. > > > (usage): Update to add -yy option. > > > * util.c (print_tracee_cwd): New function. > > > (printpathn): Update to use print_tracee_cwd and print > > > absolute path. > > > * strace.1: Add description of -yy option. > > > > I had an opportunity to think about this feature a bit longer, > > and now I see two issues I missed before. > > > > The first issue is the choice of option name. While -y option enables > > *additional* decoding of descriptors, and -yy was planned for extended > > decoding of descriptors like socket decoding, this feature *changes* > > decoding of all paths. I'd rather not mix these things. > > Sounds good, just to make sure I understand clearly -- > > 1. -y should do absolute decoding of file descriptors always.
No, -y should not change, conversion of all relative paths to absolute should be controlled by another option. > 2. -yy should do extended decoding like socket decoding etc. Yes. > > The second issue is *at syscalls. This implementation assumes that all > > non-absolute paths are relative to cwd, but that is not necessarily the > > case for *at syscalls. > > I see -- could we try and display canonical paths for fds passed into *at > syscalls? When descriptors decoding is enabled, descriptors passed to *at syscalls are decoded anyway, so we could use this information to decode relative paths passed to *at syscalls. -- ldv
pgplHswCYQ5QX.pgp
Description: PGP signature
------------------------------------------------------------------------------
_______________________________________________ Strace-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/strace-devel
