Re: [LTP] Proposed additions to runltp: run only tests matching a pattern (resend)

[Robert W. <ro...@us...>]

Thanks for the features....I applied the patch to our CVS tree.

-Robbie
(Embedded image moved to file: pic05035.jpg)

ltp...@li... wrote on 01/29/2005 12:23:29 AM:

> Attached are some revisions I'd like to submit to the Linux Test
> Project's
> runltp.  My main motive was that I wanted to be able to run a small
> number of
> tests conveniently.  The existing method, of creating a test file, was
> inconvenient for quick testing, and I encountered an apparent bug while
> using
> it (#1106474).
>      I added an option -s, which takes a string, and only runs files
> whose test case line matches the given string.  For my own use in
> testing, I
> also added a -v option, which prints the command line used to invoke
> pan.  I
> also made a couple of minor corrections to the usage text.
>      There are a number of limitations to my code for -s:
> * It relies on the presence of grep.  I don't know if the absence of
> grep is
> at all common in Linux testing, but there should be no impact on anyone
> not
> using the -s option.  (This restricted my coding significantly; it
> would have
> been more elegant to use grep everywhere, handling the absence of the
> -s flag
> as a universal pattern.)
> * I don't support the -f and -N cases in conjunction with -s; instead I
> report an
> error and exit.
> * grep searches the whole test case line; searching just the tag might
> have
> been neater, but would require relying on grep details more than seemed
> prudent.
> * -s doesn't report an error if there are no matches; pan eventually
> reports
> the problem instead, not very informatively.
> * My code doesn't check for failure to create an output file.  The
> existing
> code does this by repeatedly checking the return code from cat.  But
> grep's
> return code can be non-zero if a pattern doesn't match a particular
> file.
>
> My implementation of -v is also suboptimal; I originally used a
> variable for
> both the echoing of the command line invoking pan, and for executing
> it.  But
> when the command results in an error, the error message included my
> variable
> name rather than the command.  So the command and variable are now
> duplicated
> code.
>
> ---
> fl...@po...
> http://pobox.com/~flash
> Developer Tools Quality Assurance, PalmSource Inc.
> [attachment "runltp" deleted by Robert Williamson/Austin/IBM]
> [attachment "My_ltp_patch_2005-01-26.txt" deleted by Robert
> Williamson/Austin/IBM]