test failures on Solaris (was Re: zsh-5.2-test-2)

test failures on Solaris (was Re: zsh-5.2-test-2)

[Danek Duvall]

I have two failures on Solaris, but what seem like the obvious fixes don't
seem to work.  First:

    .../zsh.git/Test/B02typeset.ztst: starting.
    --- /tmp/zsh.ztst.28063/ztst.err        2016-12-06 14:57:07.724783240 -0800
    +++ /tmp/zsh.ztst.28063/ztst.terr       2016-12-06 14:57:07.725902052 -0800
    @@ -1 +1 @@
    -(eval):2: failed to change user ID: operation not permitted
    +(eval):2: failed to change user ID: not owner
    Test .../zsh.git/Test/B02typeset.ztst failed: error output differs from expected as shown above for:
      if (( UID )); then
	UID=$((UID+1)) date; echo "Status is printed, $?"
      else
	ZTST_skip="cannot test setuid error when tests run as superuser"
      fi
    Was testing: when cannot change UID, the command isn't run

I tried

    --- a/Test/B02typeset.ztst
    +++ b/Test/B02typeset.ztst
    @@ -720,4 +720,4 @@
     0:when cannot change UID, the command isn't run
     # 'date' did not run.
     >Status is printed, 1
    -?(eval):2: failed to change user ID: operation not permitted
    +*?(eval):2: failed to change user ID: *

But then it complained with

    -(eval):2: failed to change user ID: *
    +(eval):2: failed to change user ID: not owner

as if it wasn't expanding properly.

The second:

    .../zsh.git/Test/C02cond.ztst: starting.
    This test takes up to 60 seconds...
    --- /tmp/zsh.ztst.671/ztst.err  2016-12-06 15:09:45.098489090 -0800
    +++ /tmp/zsh.ztst.671/ztst.terr 2016-12-06 15:09:45.104223363 -0800
    @@ -0,0 +1 @@
    +(eval):1: command not found: mount
    Test .../zsh.git/Test/C02cond.ztst failed: error output differs from expected as shown above for:
    ...

I tried 

    --- a/Test/C02cond.ztst
    +++ b/Test/C02cond.ztst
    @@ -146,7 +146,7 @@
     
     # can't be bothered with -S
     
    -  if [[ ${mtab::="$({mount || /sbin/mount})"} = *[(]?*[)] ]]; then
    +  if [[ ${mtab::="$({mount || /sbin/mount || /usr/sbin/mount})"} = *[(]?*[)] ]]; then
	 print -u $ZTST_fd 'This test takes two seconds...'
       else
	 unmodified_ls="$(ls -lu $unmodified)"

but it failed with the same error.

I also see V10private.ztst, but I think that's just repeating the
B02typeset.ztst error.

Thanks,
Danek

Re: test failures on Solaris (was Re: zsh-5.2-test-2)

[Bart Schaefer]

On Dec 6,  3:33pm, Danek Duvall wrote:
} 
}     --- a/Test/B02typeset.ztst
}     +++ b/Test/B02typeset.ztst
}     @@ -720,4 +720,4 @@
}      0:when cannot change UID, the command isn't run
}      # 'date' did not run.
}      >Status is printed, 1
}     -?(eval):2: failed to change user ID: operation not permitted
}     +*?(eval):2: failed to change user ID: *

The parens in (eval) are pattern characters, so [[ "(eval):" != (eval): ]]


}     -  if [[ ${mtab::="$({mount || /sbin/mount})"} = *[(]?*[)] ]]; then
}     +  if [[ ${mtab::="$({mount || /sbin/mount || /usr/sbin/mount})"} = *[(]?*[)] ]]; then


"mount" is going to print "command not found" even if /sbin/mount is
subsequently found.  We need to throw away stderr.


diff --git a/Test/B02typeset.ztst b/Test/B02typeset.ztst
index 9c56c7e..b27bb4f 100644
--- a/Test/B02typeset.ztst
+++ b/Test/B02typeset.ztst
@@ -720,4 +720,4 @@
 0:when cannot change UID, the command isn't run
 # 'date' did not run.
 >Status is printed, 1
-?(eval):2: failed to change user ID: operation not permitted
+*?*: failed to change user ID: *
diff --git a/Test/C02cond.ztst b/Test/C02cond.ztst
index 27a2259..3852501 100644
--- a/Test/C02cond.ztst
+++ b/Test/C02cond.ztst
@@ -146,7 +146,7 @@
 
 # can't be bothered with -S
 
-  if [[ ${mtab::="$({mount || /sbin/mount})"} = *[(]?*[)] ]]; then
+  if [[ ${mtab::="$({mount || /sbin/mount || /usr/sbin/mount} 2>/dev/null)"} = *[(]?*[)] ]]; then
     print -u $ZTST_fd 'This test takes two seconds...'
   else
     unmodified_ls="$(ls -lu $unmodified)"

Re: zsh-5.2-test-3

[Bart Schaefer]

[Moving to -workers]

On Dec 10,  6:07pm, Ray Andrews wrote:
} Subject: Re: zsh-5.2-test-3
}
} On 12/09/2016 11:54 AM, Peter Stephenson wrote:
} >
} > 5.2-test-3 is in
} >   
} > http://www.zsh.org/pub/development/
} >
} >
}      Seems to have fired up ok, however:

What is the platform where you are compiling?

} 
}     builtin.c:5041:11: warning: "cursplit" may be used uninitialized in
}     this function [-Wmaybe-uninitialized]
} 
}     compmatch.c:774:15: warning: "savl" may be used uninitialized in
}     this function [-Wmaybe-uninitialized]

These are both spurious -- each of these variables shadows another which
is always used/tested first and which is not used uninitialized.  May as
well shut up the warnings, though.

}     $ make check
} 
}     ./V10private.ztst: starting.
}     Pattern match failed:
} 
} 

There ought to be more output than that ... please run

$ ZTST_verbose=2 make check TESTNUM=V10

} BTW, if there a 'make check' that only shows errors or warnings?  No 
} need to see all the stuff that's fine.

No, but as shown above you can run specific tests if you already know
the others to be OK.


diff --git a/Src/Zle/compmatch.c b/Src/Zle/compmatch.c
index f82f00e..aedf463 100644
--- a/Src/Zle/compmatch.c
+++ b/Src/Zle/compmatch.c
@@ -674,7 +674,7 @@ match_str(char *l, char *w, Brinfo *bpp, int bc, int *rwlp,
 		     *
 		     * operation.  Similar to savw.
 		     */
-		    char savl;
+		    char savl = 0;
 		    /*
 		     * The anchor on this end.
 		     */
diff --git a/Src/builtin.c b/Src/builtin.c
index e641a97..65e0cb1 100644
--- a/Src/builtin.c
+++ b/Src/builtin.c
@@ -4092,7 +4092,7 @@ bin_print(char *name, char **args, Options ops, int func)
     char *start, *endptr, *c, *d, *flag, *buf = NULL, spec[14], *fmt = NULL;
     char **first, **argp, *curarg, *flagch = "'0+- #", save = '\0', nullstr = '\0';
     size_t rcount = 0, count = 0;
-    size_t *cursplit, *splits = 0;
+    size_t *cursplit = 0, *splits = 0;
     FILE *fout = stdout;
 #ifdef HAVE_OPEN_MEMSTREAM
     size_t mcount;

Re: zsh-5.2-test-3

[Daniel Shahaf]

Bart Schaefer wrote on Sat, Dec 10, 2016 at 19:07:37 -0800:
> +++ b/Src/Zle/compmatch.c
> @@ -674,7 +674,7 @@ match_str(char *l, char *w, Brinfo *bpp, int bc, int *rwlp,
>  		     *
>  		     * operation.  Similar to savw.
>  		     */
> -		    char savl;
> +		    char savl = 0;

I see you've pushed this, but for what it's worth, it wouldn't have been
my first choice.  I'd rather have left the variable uninitialized in the
(sfx is false) case, precisely so reading savl without checking sfx
first would result in an "is used uninitialized" warning.

Re: zsh-5.2-test-3

[Bart Schaefer]

On Dec 11,  3:26am, Daniel Shahaf wrote:
} Subject: Re: zsh-5.2-test-3
}
} Bart Schaefer wrote on Sat, Dec 10, 2016 at 19:07:37 -0800:
} > -		    char savl;
} > +		    char savl = 0;
} 
} I see you've pushed this, but for what it's worth, it wouldn't have been
} my first choice.  I'd rather have left the variable uninitialized in the
} (sfx is false) case, precisely so reading savl without checking sfx
} first would result in an "is used uninitialized" warning.

Obviously though at least one compiler isn't clever enough to spot that
detail, otherwise Ray wouldn't have seen the warning to begin with.

Re: zsh-5.2-test-3

[Ray Andrews]

[-- Attachment #1: Type: text/plain, Size: 4219 bytes --]

On 12/10/2016 07:07 PM, Bart Schaefer wrote:
> What is the platform where you are compiling?

Debian testing in a rather old state of update
> }     $ make check
> }
> }     ./V10private.ztst: starting.
> }     Pattern match failed:
> }
> }
>
> There ought to be more output than that ... please run
>
> $ ZTST_verbose=2 make check TESTNUM=V10
Verbose is right!:

    9 /aMisc/Zsh-new/zsh-5.2-test-3 2$ ZTST_verbose=2 make check TESTNUM=V10
    cd Test ; make check
    make[1]: Entering directory '/aMisc/Zsh-new/zsh-5.2-test-3/Test'
    if test -n "gcc"; then \
       cd .. && DESTDIR= \
       make MODDIR=`pwd`/Test/Modules install.modules > /dev/null; \
    fi
    if ZTST_testlist="`for f in ./V10*.ztst; \
                do echo $f; done`" \
      ZTST_srcdir="." \
      ZTST_exe=../Src/zsh \
      ../Src/zsh +Z -f ./runtests.zsh; then \
      stat=0; \
    else \
      stat=1; \
    fi; \
    sleep 1; \
    rm -rf Modules .zcompdump; \
    exit $stat
    ./V10private.ztst: starting.
    ZTST_getsect: read section name: prep
    ZTST_getchunk: read code chunk:
      if ! zmodload zsh/param/private 2>/dev/null; then
        ZTST_unimplemented="can't load the zsh/param/private module for
    testing"
      else
        # Do not use .tmp here, ztst.zsh will remove it too soon (see
    %cleanup)
        mkdir private.TMP
        sed -e 's,# test_zsh_param_private,zmodload zsh/param/private,'
    < $ZTST_srcdir/B02typeset.ztst > private.TMP/B02
      fi
    ZTST_execchunk: status 0
    ZTST_getchunk: read code chunk:

    ZTST_getsect: read section name: test
    ZTST_test: looking for new test
    ZTST_test: examining line:

    ZTST_test: examining line:
      (zmodload -u zsh/param/private && zmodload zsh/param/private)
    ZTST_getchunk: read code chunk:
      (zmodload -u zsh/param/private && zmodload zsh/param/private)
    ZTST_test: examining line:

    Running test: unload and reload the module without crashing
    ZTST_test: expecting status: 0
    Input: /tmp/zsh.ztst.8681/ztst.in, output:
    /tmp/zsh.ztst.8681/ztst.out, error: /tmp/zsh.ztst.8681/ztst.terr
    ZTST_execchunk: status 0
    ZTST_test: test produced standard output:

    ZTST_test: and standard error:

    Test successful.
    ZTST_test: looking for new test
    ZTST_test: examining line:

    ZTST_test: examining line:
      ZTST_verbose=0 $ZTST_exe +Z -f $ZTST_srcdir/ztst.zsh private.TMP/B02
    ZTST_getchunk: read code chunk:
      ZTST_verbose=0 $ZTST_exe +Z -f $ZTST_srcdir/ztst.zsh private.TMP/B02
    ZTST_test: examining line:
    *>*
    ZTST_getredir: read redir for '>':
    *
    ZTST_test: examining line:
    *>*
    ZTST_getredir: read redir for '>':
    *
    ZTST_test: examining line:

    Running test: typeset still works with zsh/param/private module loaded
    ZTST_test: expecting status: 0
    Input: /tmp/zsh.ztst.8681/ztst.in, output:
    /tmp/zsh.ztst.8681/ztst.out, error: /tmp/zsh.ztst.8681/ztst.terr
    ZTST_execchunk: status 0
    ZTST_test: test produced standard output:
    private.TMP/B02: starting.
    Test case skipped: cannot test setuid error when tests run as superuser
    private.TMP/B02: all tests successful.
    ZTST_test: and standard error:

    Pattern match failed:
    <*
    <*
     >private.TMP/B02: starting.
     >Test case skipped: cannot test setuid error when tests run as
    superuser
     >private.TMP/B02: all tests successful.
    Test ./V10private.ztst failed: output differs from expected as shown
    above for:
      ZTST_verbose=0 $ZTST_exe +Z -f $ZTST_srcdir/ztst.zsh private.TMP/B02
    Was testing: typeset still works with zsh/param/private module loaded
    ./V10private.ztst: test failed.
    ZTST_getsect: read section name: clean
    ZTST_getchunk: read code chunk:
       rm -r private.TMP
    ZTST_execchunk: status 0
    ZTST_getchunk: read code chunk:

    **************************************
    0 successful test scripts, 1 failure, 0 skipped
    **************************************
    Makefile:187: recipe for target 'check' failed
    make[1]: *** [check] Error 1
    make[1]: Leaving directory '/aMisc/Zsh-new/zsh-5.2-test-3/Test'
    Makefile:263: recipe for target 'check' failed
    make: *** [check] Error 2

Re: zsh-5.2-test-3

[Bart Schaefer]

On Dec 10,  7:26pm, Ray Andrews wrote:
} 
}     Running test: typeset still works with zsh/param/private module loaded
}     ZTST_test: expecting status: 0
}     Input: /tmp/zsh.ztst.8681/ztst.in, output:
}     /tmp/zsh.ztst.8681/ztst.out, error: /tmp/zsh.ztst.8681/ztst.terr
}     ZTST_execchunk: status 0
}     ZTST_test: test produced standard output:
}     private.TMP/B02: starting.
}     Test case skipped: cannot test setuid error when tests run as superuser

Well, there you go.  Stop running "make check" while superuser.

I'm not sure this one is worth attempting to fix.  The output of B02
is different for EUID=0 and V10 is attempting to pattern-match the
expected successful output of B02.  We'd need two different test cases
with different output comparisons depending on whether "make check" is
being run as root.

Actually -- why is that test in B02 in the first place, rather than in
say D04parameter.ztst ?

Re: zsh-5.2-test-3

[Ray Andrews]

On 12/10/2016 07:44 PM, Bart Schaefer wrote:
> Well, there you go.  Stop running "make check" while superuser.

Yabut there's a few other things that are very specific that they won't 
run as superuser, so you know you needn't worry about them as the 
messages are careful to explain.  Shouldn't this one be similar?  As for 
me, I'm always super user, I got sick of typing sudo 500 times a day 
;-)  They say we're not supposta, but it's my computer and I'll bust it 
if I wanta.  Which I do,  and there's always a lesson to learned from it.

Just to  be kosher:

     $ su ray
     $ make check
...

**************************************
47 successful test scripts, 0 failures, 1 skipped
**************************************
make[1]: Leaving directory '/aMisc/Zsh-new/zsh-5.2-test-3/Test'
ray@HP-w5--8-Debian1:/aMisc/Zsh-new/zsh-5.2-test-3$

... I'll remember that for next time, thanks Bart.  Dunno, maybe there 
could be a warning about that before the rest of the test runs?

Re: zsh-5.2-test-3

[Bart Schaefer]

On Dec 10,  7:59pm, Ray Andrews wrote:
} 
} Yabut there's a few other things that are very specific that they won't 
} run as superuser, so you know you needn't worry about them as the 
} messages are careful to explain.  Shouldn't this one be similar?

Ideally yes, but in this case we're re-running B02 from inside V10;
B02 got changed without V10 being changed, so my question is, where
do we fix it, or do we just not bother because running the tests as
superuser is so unusual in the first place.

Re: zsh-5.2-test-3

[Ray Andrews]

[-- Attachment #1: Type: text/plain, Size: 862 bytes --]

On 12/11/2016 01:36 AM, Bart Schaefer wrote:
> On Dec 10,  7:59pm, Ray Andrews wrote:
> }
> } Yabut there's a few other things that are very specific that they won't
> } run as superuser, so you know you needn't worry about them as the
> } messages are careful to explain.  Shouldn't this one be similar?
>
> Ideally yes, but in this case we're re-running B02 from inside V10;
> B02 got changed without V10 being changed, so my question is, where
> do we fix it, or do we just not bother because running the tests as
> superuser is so unusual in the first place.
>
Not mine to say of course, I suspect that a message would be easy 
enough, but any complexities are to be avoided. I'm a spit and polish 
fanatic, polish it until it gleams. Not the culture in unix/linux tho -- 
I suppose something  so  huge can never be polished like we polished 
things in DOS.

Re: zsh-5.2-test-3

[Bart Schaefer]

On Dec 11,  9:01am, Ray Andrews wrote:
}
} I suppose something  so  huge can never be polished like we polished 
} things in DOS.

Hey, if somebody wants to pay me to polish zsh (warning, I am not
inexpensive) like everyone who polished DOS got paid, let me know
and we can try to work something out.

Re: zsh-5.2-test-3

[Daniel Shahaf]

Bart Schaefer wrote on Sat, Dec 10, 2016 at 19:44:21 -0800:
> Actually -- why is that test in B02 in the first place, rather than in
> say D04parameter.ztst ?

Because it's about assignment, as opposed to substitution?  If there's
a better home for the test I don't mind moving it.

(will reply to your other mail later or tomorrow)

Re: zsh-5.2-test-3

[Bart Schaefer]

On Dec 11,  4:11am, Daniel Shahaf wrote:
} Subject: Re: zsh-5.2-test-3
}
} Bart Schaefer wrote on Sat, Dec 10, 2016 at 19:44:21 -0800:
} > Actually -- why is that test in B02 in the first place, rather than in
} > say D04parameter.ztst ?
} 
} Because it's about assignment, as opposed to substitution?

Then why isn't in it A06assign ?  (Because it's not really about
assignment, either -- it's about a special-parameter side-effect.)

} If there's a better home for the test I don't mind moving it.

It probably ought to be in A05execution, or maybe we need a new test
file just for behavior of assorted special parameters.  C06specials ?

Incidentally, assigning to USERNAME attempts to change both UID and
GID and is still only a warning, not an error.  Further, if setuid()
and setgid() aren't available, assignments to UID et al. silently
do nothing, which is probably why failure of set?id() when available
was never a hard error before.

Passing thought, mostly unrelated:  There should be tests of "getopts"
in B08shift.  I believe we still have some incompatibility with POSIX
in OPTIND/OPTARG handling it would be nice to have an example of the
current behavior with some mention of the difference in case we get
around to fixing it.