William A. Rowe, Jr. wrote:
>I don't see any issues with your suggestions below. I'll sanity check
>your commit after you've finished (if you haven't already) :-)
>
It's done already. No external changes, just the values in the win32
apr_arch_file_io.h changed, and I added a note to apr_file_io.h about
the reserved range.
Brane
>Bill
>
>At 08:01 PM 7/3/2003, Branko ÃÅibej wrote:
>
>
>>In include/arch/win32/apr_arch_file_io.h, we have the following definitions:
>>
>>/* Internal Flags for apr_file_open */
>>#define APR_OPENINFO 0x4000 /* Open without READ or WRITE access */
>>#define APR_OPENLINK 0x2000 /* Open a link itself, if supported */
>>#define APR_READCONTROL 0x1000 /* Read the file's owner/perms */
>>#define APR_WRITECONTROL 0x0800 /* Modifythe file's owner/perms */
>>
>>I see a couple of problems here:
>>
>> * APR_READCONTROL and APR_WRITECONTROL are in conflict with
>> APR_SENDFILE_ENABLED and APR_FILE_NOCLEANUP, which are defined in
>> include/apr_file_io.h.
>> * These two flags are never actually checked for in apr_file_open.
>> Why are they defined at all?
>>
>>I'd suggest moving the values of these internal flags into the
>>0x00100000--0x80000000 range, and adding a note to apr_file_io.h that
>>those values are reserved for internal flags. Note that the flags
>>parameter is an apr_int32_t, so that range is always available.
>>
>>Right now, APR_READCONTROL can get passed to apr_file_open from
>>apr_stat, but is then interpreted as APR_SENDFILE_ENABLED ... weird, to
>>say the least.
>>
>>Thoughts? If nobody objects, I'll make that change soonish. Somebody
>>should probably check the other platforms for similar overlap,
>>
>>--
>>Brane ÃÅibej <[EMAIL PROTECTED]> http://www.xbc.nu/brane/
>>
>>
>
>
>
>
--
Brane Äibej <[EMAIL PROTECTED]> http://www.xbc.nu/brane/
