boost::filesystem::exists throws on Windows if the drive letter doesn't exist

[davidjward30@…]

boost::filesystem::exists throws on Windows if the drive letter doesn't exist.

This is rather unexpected. It's rather tedious to validate the drive letter beforehand or to have to catch the exception.

I think it's reasonable to say that if the drive doesn't exist then the file doesn't exist.

I noted that status_template checks a list of errors returned by get_file_attributes:

if ((ec.value() == ERROR_FILE_NOT_FOUND)

(ec.value() == ERROR_PATH_NOT_FOUND)

(ec.value() == ERROR_INVALID_NAME) "tools/jam/src/:sys:stat.h", "foo"

(ec.value() == ERROR_INVALID_PARAMETER) ":sys:stat.h"

(ec.value() == ERROR_BAD_PATHNAME) "nosuch" on Win64

(ec.value() == ERROR_BAD_NETPATH

I would propose adding ERROR_INVALID_DRIVE to this list.

[anonymous]

Hello,

I have a correction to this issue. My invalid drive is actually a removable disk (USB) so actually the drive letter *does* exist but nothing is inserted. I suppose its therefore justified that an exception is thrown.

Yours,

David.

[megibson@…]

If the drive does not exist, then the file does not exist. Special cases are nice, but this function does not do the right thing when the drive does not exist. I don't see how the previous comment is relevant.

[Beman Dawes]

(In [51931]) Fix #2725

[claudio@…]

boost::filesystem::exists also throws if the drive is an empty CD/DVD drive.

I think it's reasonable to say that if the CD/DVD drive is empty then the file doesn't exist.

For an empty CD/DVD drive ec.value() == ERROR_NOT_READY

[Beman Dawes]

I can't reproduce the problem with v3 on the current trunk.

I've added a program, .../libs/filesystem/v3/example/file_status.cpp to the trunk. It can be built with bjam in that directory.

Please give it a try and send me the output if you are still having problems.

--Beman