On 20/04/2011 17:36, Corinna Vinschen wrote: > On Apr 8 15:43, Jon TURNEY wrote: >> At the moment, all errors in archive::extract_file() are assumed to be due >> to a failure to open the file for writing due to it being opened by another >> process. >> >> Distinguish when the error is due to an inability to read the file from the >> source archive, and report that. >> >> 2011-04-08 Jon TURNEY <...> >> >> * install.cc (extract_replace_on_reboot): New function containg code >> extracted from... >> (installOne): Report read errors differently to write errors >> * archive.cc (extract_file): Distinguish read errors from write errors > > This loooks good to me. Just a question... > >> + /* XXX FIXME: prefix may not be / for in use files - >> + * although it most likely is >> + * - we need a io method to get win32 paths >> + * or to wrap this system call > > This comment doesn't make any sense to me anymore. The incoming path > should already be a full Win32 path. The function wouldn't work without > it and AFAIK the paths from the tar file are converted to full Win32 > paths anyway. There's also no reason to wrap this system call any further. > Bottom line is, just drop the comment, please.
Ok. I've applied patches 1-4, with 4 amended as suggested here. For convenience, I have generated a snapshot at [1] (Note that this was built with Charles Wilson's i686-pc-mingw32-gcc 4.5.1 toolchain, as that's what I have installed) [1] http://cygwin.com/setup/snapshots/setup-2.745.exe
