On Sat, Dec 18, 2004 at 04:12:17PM -0500, Igor Pechtchanski wrote: >On Fri, 17 Dec 2004, Christopher Faylor wrote: > >> On Fri, Dec 17, 2004 at 06:01:04PM -0500, Pierre A. Humblet wrote: >> >Christopher Faylor wrote: >> > >> >> While I detest the trailing dot crap, I don't want cygwin to be >> >> inconsistent. I don't want ls /bin./ls.exe to fail but ls >> >> /cygdrive/c/bin./ls.exe to work. >> > >> >Assuming a normal install, the first one is c:\cygwin\bin.\ls.exe, >> >which would NOT fail, while the second is c:\bin.\ls.exe, which would >> >fail as expected (not due to dots). >> >> Ok. Yes. I had a typo. >> >> If /cygdrive/c/cygwin/bin./ls.exe works, then /bin./ls.exe should also >> work. Or, both should fail. "consistent" > >If I may chime in, I think there are at least three separate possibilities >for accessing each directory: > >1) via a managed Cygwin mount; >2) via a regular Cygwin mount; >3) via a /cygdrive-prefixed path; and maybe >4) via a Win32 path. > >There's a need for consistency in each of the above cases, but not >between cases. Each may justifiably have different behavior. We >already default to textmode for 4), and 2) and 3) may have different >textmode/binmode behavior.
2 and 3 don't have different textmode/binmode behavior. You use mount to control the behavior. >It could be argued that as you go down this list, the POSIXness >decreases, so it's ok to distinguish trailing dots, e.g., in the first >two cases, and ignore them in the latter two. So, you're saying that 2 should always say "file not found"? Non-managed mounts can't reliably distinguish trailing dots. Unless you want to make every file a symbolic link to the "real" file or always issue an ENOENT when a pathname component has a trailing dot. I don't see a difference between /cygdrive or a mounted drive. Both are cygwin inventions. I don't see a logical explanable reason for what they support to be different. That is another case that I didn't specify. Should /cygdrive./c work? I think I'm coming back to my ENOENT for all trailing dots argument... >Just my $0.02. Using past behavior, which has certainly changed in the last six months and undoubtedly changed a couple of times prior to that, or extrapolating from the behavior of textmode/binmode to justify this is really not an argument that I'm interested in. You can cause cygwin programs to control the behavior of textmode/binmode with appropriate flags to fopen or open or by specifying the explicit use of mount. There is no easy way for a cygwin program to figure out what to do with trailing dots. cgf
