[Toybox] [Bug] - single command mode and testsuite

[Rob Landley]

On 11/26/2013 05:55:41 AM, Ashwini Sharma wrote:
> Hi Rob, List,
> 
> single command build:-
> 
> While doing a single command build for the commands which are aliased
> (OLDTOY)
> of some other command, SIGSEGV is reported.
> 
> This is due to the fact that OLDTOY doesn't have the CONFIG_ macro for
> itself.

Yeah, I need to upgrade scripts/single.sh to build those properly.

> So while doing the single command build for such a command, only TEST
> command is there in toy_list. which causes a Segfault.

Huh, I was getting build breaks here last time I tried to build an  
OLDTOY with single.sh. (The build was generating files with missing  
data.) Let's see...

$ scripts/single.sh mv
scripts/make.sh
Make generated/config.h from .singleconfig.
Extract configuration information from toys/*.c files...
Generate headers from toys/*/*.c...
generated/newtoys.h
generated/flags.h
generated/globals.h
generated/help.h
Extract help text from Config.in.
Library probe...
Compile toybox...
main.c: In function 'toy_find':
main.c:37:3: warning: overflow in implicit constant conversion  
[-Woverflow]
main.c: In function 'main':
main.c:176:13: warning: array subscript is above array bounds  
[-Warray-bounds]
lib/help.c:16:1: error: expected expression before ';' token
lib/help.c:14:14: warning: 'help_data' defined but not used  
[-Wunused-variable]
make: *** [toybox] Error 1

TEST command? Do you mean toys/pending/test.c?

Part of the point of the single builds is to make the toy_list drop out  
and become a single structure instance, so it doesn't really as much  
what's in there because that's the only command.

> e.g. -  _mv_ is an OLDTOY of _cp_. do a single build for mv.
> run mv with arguments, i.e. ./mv
> Segfault.
> 
> Testsuite:
> blkid test case fails for squashfs, reason being Wrong number of  
> arguments.
> Which causes the testsuite to exit.

The one after it (vfat) is broken too.

Looks like I checked in a half-finished file. Lemme at least file down  
the rough edges...

Ah yes, I got distracted by switching "murderfs" back to "reiserfs".  
(Didn't mean to check that in.)

> I guess either the testcase or the Exit condition need to be reviewed.

I addition to providing the extra blank arguments, I needed to add the  
newlines to the expected output.

Thanks for the heads up,

Rob
 1387805956.0