[cvsnt] Exec failed: No such file or directory

[Michael Wojcik]

> From: Greg Lee [mailto:glee-list at swspec.com] 
> Sent: Wednesday, 22 March, 2006 10:55
> 
> 
> > Please remember to provide both client and server platform 
> 
> You are correct, Linux is at both the client (kernel version 
> 2.6.12) and server (kernel version 2.4.21) ends.

Thanks.

> > > I am having a problem when I commit changes to any file.  The 
> > > client always returns a
> > > message saying "Exec failed: No such file or directory" ...
> >
> > Offhand it appears that you should be getting more verbose logging
of
> > what CVSNT is trying to run; I'm not sure why you aren't, but it's
late
> > and I haven't looked at that part of the source.  In fact, with
-tttt I
> > get a *lot* more trace output from "cvs commit" than what you 
> > show - did you remove part of the trace output?
> 
> No, I did not remove part of the output.  I did use the form 
> "-t -t -t -t" instead of
> "-tttt", but I just tried "-tttt" and it returned the same results.

Huh.  Maybe tracing is less verbose in the 2.5 sources; I'm running 2.0
here.

> I determined that the error is coming from the server side so 
> I added some code on the
> server side to see what is being executed:
> 
> [glee at dev1 zzz]$ cvs -tttt commit -m "testing"
>  -> main loop with CVSROOT=glee at 192.168.0.94:/home/cvsroot
>  -> Starting server: /usr/bin/ssh -l glee 192.168.0.94 cvs server
> glee at 192.168.0.94's password:
>  -> Sending file `z.c' to server
> Checking in z.c;
> /home/cvsroot/zzz/z.c,v  <--  z.c
> new revision: 1.17; previous revision: 1.16
> done
>  -> unlink_file(CVS/Base/z.c)
>  -> Register(z.c, 1.17, Wed Mar 22 15:35:14 2006, ,  )
>  -> rename(CVS/Entries.Backup,CVS/Entries)
>  -> unlink_file(CVS/Entries.Log)
> Exec failed ("(echo ; id; echo zzz z.c,1.16,1.17; date; cat) >>
> /home/cvsroot/CVSROOT/commitlog "): No such file or directory
>  -> Lock_Cleanup()
> 
> This is what I did to get this information:
> 
> --- cvsnt-2.5.03.2260/cvsapi/unix/RunFile.cpp   2006-03-22 
> 10:34:54.000000000 -0500
> +++ cvsnt-2.5.03.2260.orig/cvsapi/unix/RunFile.cpp      
> 2006-01-11 13:00:35.000000000
> -0500
> @@ -184,18 +184,8 @@
> 
>                 char **myargv = (char**)m_args->toArgv();
>                 execvp (myargv[0], myargv);
> -{
> -       char buf[2000] = "Exec failed (\"";
> -       int i;
> -       for(i=0; myargv[i] != NULL && i < 20; i++)
> -       {
> -               strcat(buf, myargv[i]);
> -               strcat(buf, " ");
> -       }
> -       strcat(buf, "\")");
> -               perror(buf);
> +               perror("Exec failed");
>                 exit(-1);
> -}
>         }
> 
>         if(m_inFd>=0)

Looks OK.

> So it looks like something in "(echo ; id; echo zzz 
> z.c,1.16,1.17; date; cat)" is not
> executing.  In fact, isn't execvp trying to execute the first 
> argument in the list, i.e.
> "(echo"?  "(echo" is not valid, the "(" is wrong.  "echo" is 
> valid assuming it is in the
> path.  Maybe the intent was to pass this string to a shell 
> (sh, bash, etc)?

That would be my guess.  Unfortunately this doesn't seem to be in the
2.0.51d source, which is all I have downloaded at the moment.  Perhaps
the problem occurs if $SHELL isn't set in the server process's
environment?  myargv is constructed from m_args, which is probably built
as a command line, ie as a single string, so if the command name were
missing I bet you'd end up with the first parameter in myargv[0].

-- 
Michael Wojcik
Principal Software Systems Developer, Micro Focus