Re: Reading /proc/registry/... returns extra char

[Christopher Faylor <cgf-use-the-mailinglist-please at cygwin dot com>]

On Mon, May 31, 2010 at 03:35:19AM +0400, Andrey Repin wrote:
>Greetings, Christopher Faylor!
>>>>> >> > $ cat >a.dat
>>>>> >> > /proc/registry/HKEY_LOCAL_MACHINE/SYSTEM/CurrentControlSet/Control/Syst
>>>>> >> > emBootDevice
>>>>> 
>>>>> > This trailing NUL character was always there, already with Cygwin 1.5.
>>>>> > It's part of the file content.  If strings are stored with a trailing
>>>>> > NUL in a file, you don't want Cygwin to remove it for you, right?
>>>>> 
>>>>> Wrong. The training NULL is a string value terminator for REG_SZ variables,
>>>>> also a string separator for REG_MULTI_SZ ones. (Which ends with a spare NULL)
>>>>> It must not be exposed to the user.
>>>
>>>> I disagree.  When you're using tools like regtool, you're right.  But
>>>> when accessing the registry as *files* via the virtual /proc filesystem,
>>>> you want the file content.
>>>
>>>Yep. And I certainly not expect the NULL in text files. You know, not every
>>>console program is binary-safe when working with STDIO? Not even cygwin, as we
>>>can see in this thread.
>>>
>>>> And the file contains the trailing NUL in REG_SZ and REG_EXPAND_SZ values,
>>>> and multiple NULs in REG_MULTI_SZ values.
>>>
>>>That's right and true, when you're working with interface directly, but... all
>>>the programs I've used in the past, and all the interfaces, they do not expose
>>>trailing NULL to the client application.
>>>In this case, /proc/registry is an interface, but cat is the application.
>
>> NAME
>>        cat - concatenate files and print on the standard output
>
>> I don't see anything in cat's description which claims it should know
>> that some files are special and should be handled differently.
>
>I didn't said about cat, but rather about file it reading from.
>
>>>> What do you suppose Cygwin should do with the NULs in REG_MULTI_SZ values?
>>>> Just remove them?
>>>
>>>Convert them to appropriate EOL sequences. And back to NULL's on write. As per
>>>definition of a text as "multiple strings".
>>>You don't need to argue over it, just document it properly :)
>
>> Converting a NUL to a EOL would be very strange behavior.
>
>Strange? I don't think so. I expect text data from text file. Not binary
>stream.
>
>> I don't think you really know what you're asking for.
>
>I know. Look at windows own regedit and Registry Browser FAR plugin as two
>examples.
>They do such two-way conversion transparently for user.

Ok.  I'm done trying to educate.

We're not going to be changing Cygwin unless we find that there is a bug
in its representation of the data in /proc/registry.  If the data is
being exactly represented as it is retrieved from the registry API, it
is not a bug.

Even if there is a bug, we are not going to be translating NUL to EOL.
I'm fairly confident that Corinna will back me up on that.

cgf

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple