Re: Directory masks -- howto?

Hello,

I finally understood the problem and how this bug lived since version 1.2.0

yes, this is a bug in libdar 2.0.0 pre1. In version 1.3.0 the warning 
code is commented out by mistake (I forgot to uncomment it) thus the 
message is never shown. in 1.2.1 and before a bug (fixed in 1.3.0) made 
the warning appear in very rare cases (the fix was in mask.cpp from 
version 1.10 to 1.11, but the code in dar.cpp was kept commented out) so
the bug you've found never shown up to today.

OK, so the solution is as you said, there is no need to have

       && subtree.is_covered((sauv_path +
       (filename+".1."+extension)).display()))

but instead we need

      && subtree.is_covered(sauv_path.display()))

I update CVS, and will soon make pre-release 2. (I must check few other 
reported problem first).

Thanks Johnathan !
By the way, your dar_test program was very useful to test the 
distribution usability... it helped me se some problem about the 
installation concerning include files ... thanks again !

Regards,
Denis.





Johnathan Burchill wrote:
> On October 14, 2003 02:36 pm, you wrote:
> [...]
> 
>>Strange. This warning is issued only if
>>1 - you create an archive
>>*and*
>>2 - the filename mask dot exclude the first slice filename
>>*and*
>>3 - the directory where is saved the archive (sauv_root) is not excluded
>>by the subtree filter (subtree).
>>
>>1 is true
>>2 is true
>>for the warning not to appear 3 must thus be false.
>>this lead me to the question, which arguments do you give to the
>>op_extract() call, in particular, the fs_root and sauv_root ?
>>
> 
> Just to be clear, I am doing an "op_create", and I've now hard-coded the 
> fs_root and sauv_root into the op_create call to eliminate any potential 
> problems with my filename parsing.
> 
> So here is an example file ("dar-test-create") which reproduces my problem.
> 
> #include <iostream>
> #include <string>
> #include <libdar.hpp>
> #include <erreurs.hpp>
> 
> using namespace std;
> using namespace libdar;
> 
> void interaction_warning (const std::string &);
> bool interaction_pause (const std::string &);
> 
> int main ( )
> {
>         set_warning_callback(&interaction_warning);
>         set_answer_callback(&interaction_pause);
> 
> 	user_interaction_warning("Basic program working");
> 	try
> 	{
> 		user_interaction_pause("About to create the archive. Do you wish to 
> continue?");
> 	}
> 	catch (Euser_abort &e)
> 	{
> 		user_interaction_warning(e.get_message());
> 		return -1;
> 	}
> 	try
> 	{
> 	op_create("/home/john/test-dar-create", 
> "/home/john/test-dar-create/backups",
> 		NULL, bool_mask(true), not_mask( simple_mask( 
> "/home/john/test-dar-create/backups")),
> 		"test-archive", "dar", NULL, true, true, true, true,
> 		true, bzip2, 9, 650000000, 645000000, true, true,
> 		"", "", "", "", "", "", bool_mask(true), 0, true,
> 		false, 0);
> 	}
> 	catch (Egeneric & e)
> 	{
> 		user_interaction_warning(e.get_message());
> 	}
> }
> 
>  void interaction_warning(const std::string &message)
> {
> 	cout << message.c_str();
> }
>  bool interaction_pause(const std::string &message)
> {
> 	string answer;
> 	cout << message.c_str();
> 	cin >> answer;
> 	if (answer == "y")
> 	{
> 		return true;
> 	}
> 	else
> 	{
> 		return false;
> 	}
> }
> 
> 
> I compiled it with...
> g++ -DHAVE_CONFIG_H -fexceptions -ldar64 -DMODE=64 -I./dar/src/libdar -o 
> dar-test-create dar-test-create.cpp
> 
> I get the "recursive error" message when running "dar-test-create".
> The archive gets created anyway without any problem though.
> 
> Here is the output from the created archive:
> bash-2.05b$ dar -l test-archive.1.dar
> Reading config file: /home/john/.darrc
> Warning, test-archive.1.dar seems more to be a slice name than a base name. 
> Do you want to replace it by test-archive ? [return = OK | esc = cancel]
> Continuing...
> [data ][ EA  ][compr] | permission | user  | group | size  |          date                 
> |    filename
> ----------------------+------------+-------+-------+-------+-------------------------------+------------
> [Saved]       [-----]   drwxr-xr-x   john       users   0       Wed Oct 15 
> 11:06:27 2003  backups
> [Saved]       [  67%]   -rwxr-xr-x   john       users   1239391 Wed Oct 15 
> 10:57:42 2003  kdar
> 
> So dar does exclude the backup directory, but the warning message gets 
> issued anyway.
> 
> The libdar code which generates the "recursion error" message is:
> 
> 	    if(create_not_isolate && sauv_path.is_subdir_of(fs_root)
> 	       && selection.is_covered(filename+".1."+extension)
> 	       && subtree.is_covered((sauv_path + 
> (filename+".1."+extension)).display()))
> 		user_interaction_pause(string("WARNING! The archive is located in the 
> directory to backup, this may create an endless loop when the archive will 
> try to save itself. You can either add -X \"")+ filename + ".*." + 
> extension +"\" on the command line, or change the location of the archive 
> (see -h for help). Do you really want to continue?");
> 
> What I don't understand here is that if subtree is supposed to mask 
> directories, why does subtree.is_covered() include a filename? Shouldn't 
> there just be "sauv_path" as an argument?
> 
> 
>>dar also calls op_extract, and in the current situation would call
>>something like this:
>>
>>	op_extract("/home/john/kdar", "/home/john/kdar/backups",
>>	bool_mask(true), subtree, ....)
>>
>>with subtree built as you did:
>>	ou_mask exclus;
>>	exclus.add_mask(simple_mask("/home/john/kdar/backups"));
>>	et_mask subtree;
>>	subtree.add_mask(not_mask(exclus));
>>
>>(you could also replace "subtree" by "not_mask(exclus)" in the argument
>>given to op_extract, it should do the same in this case.
>>
>>
>>>It works fine when I have two "-P" type arguments:
>>>exdir1="/home/john/kdar/backups"
>>>exdir2="/home/john/kdar/backups/*"
>>>exclus.add_mask(simple_mask(exdir1));
>>>exclus.add_mask(simple_mask(exdir2));
>>
>>this should not be necessary.
>>
>>
>>>The command line version of libdar does not require me to have both
>>>"-P" arguments for the backup to be excluded, I just need "-P
>>>backups". How do I make it as simple using libdar directly?
>>
>>we will find out, but be sure that dar command line uses the same
>>interface as kdar and no such trick is necessary to avoid the warning
>>message.
> 
> 
> Absolutely.
> 
>>>Thanks,
>>>JB
>>>
>>>P.S. some screenshots of KDar can be found at
>>>http://www.phys.ucalgary.ca/~burchill/KDAR/index.html
>>
>>that seems very promising !!! I'm impatient to try it !  :-)))
>>
> 
> 
> I'm happy to share. I could put a tarball on my website once I have the 
> masks working. It would be good to get feedback on how to improve the UI.
> 
> JB




-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
SourceForge.net hosts over 70,000 Open Source Projects.
See the people who have HELPED US provide better services:
Click here: http://sourceforge.net/supporters.php