Re: [MiNT] XaAES: bug(s) in shel_write()

["Helmut Karlowski" <helmut.karlowski@ish.de>]

Am 05.02.2016, 03:04 Uhr, schrieb Thorsten Otto:

> > What is SHD_ENVIRON? I've found
> Its the same as SW_ENVIRON. I took the name from the XaAES sources (the  
> ones that are part of MiNT).
>
> > What exactly did you write in ARGV?
> The arguments that should be passed to the program of course, including  
> the command name. That part works,

Actually I don't think it matters what you put in ARGV.

What do you mean with:

> i even verified it again when that bug first appeared.

What XaAES-version does what you want?

> > Looks like that is not supported by XaAES.
> The code look like it tries to do so, but fails to do it correctly. If

Where does XaAES parse the ARGV-part of the caller's environment if you  
meant that.

> the original arguments are too long to be passed in the basepage, and  
> this fact is mentioned as 127 being used as length, the function should  
> either fail with an error indicating that this is not supported, or  
> support it correctly. It does not make sense to pass wrong, truncated  
> arguments to the program instead, this can be rather disastrous. BTW  
> XaAES does indicate in appl_getinfo(10) that ARGV *is* supported.

I tried with a test-program and it indeed truncated the command-line to  
124 bytes.

When I remove:

      if (longtail > 124)
        longtail = 124;

all arguments are passed to the launched program. That looks wrong.

> > Only if SW_ENVIRON is set.
> But this is the only way to pass the environment in the call that  
> contains the ARGV element. The application itself does not create a new  
> process where this could be done, nor does XaAES has access to the  
> process' environment.

I guess the purpose SW_ENVIRON is to pass the caller's environment to the  
callee's without interfering of the AES. You don't need it I think. The  
meaning of "use ARGV" on the other hand is that the AES writes the  
command-line into the environment for the callee, not that it reads it  
from there.

> > Do you have a simple test-case?
> Not really. It has to be a complete GEM program after all. I can try to  
> make one if it is really needed. But i hope my explanations made clear  
> that something is going wrong there. Not using the environment that was  
> constructed by make_argv() is at least definitely a bug. And that code

I uses it:

       ret = create_process(cmd, t,
                 (x_mode & SW_ENVIRON) ? x_shell.env : *strings,
                 &p, 0, cpopts.mode ? &cpopts : NULL);

strings is written in make_argv().

I have attached my test-programs. Is it this what you mean?

-Helmut

Attachment: callee.c
Description: Binary data

Attachment: caller.c
Description: Binary data