Martin Sebor wrote:
> Also, I noticed some strange long --trace output in a few cases that
doesn't look right to me (see below). Could you look into it to see
what's going on? (Is it because we're not using the right    >
formatting directive?)

Yes, the %{#*s} directive is used in the test for both char and wchar_t
arrays. But the %{#*ls} should be used for wchar_t.
I think the rw_assert call should be modified using the rw_narrow()
functions to avoid additional "switches" in the formatting line. Will
something like this:

...
    bool success = 
        !TestString::traits_type::compare (exp_res, s_res, exp_res_len);

    static char tmp1[long_string_len];
    static char tmp2[long_string_len];

    rw_assert (success, 0, cs.line,
               CALLFMAT " expected %{#*s}, got %{#*s}", CALLARGS,
               int (exp_res_len), rw_narrow (tmp1, exp_res,
exp_res_len),
               int (res), rw_narrow (tmp2, s_res, res));
...

be Ok?

Martin Sebor wrote:
> exp_exceptions[] was missing element[2] (dereferenced later in the
test). I added it.
I think the catch for the length_error should be removed at all - this
exception isn't expected here.

Thanks,
Anton Pevtsov

-----Original Message-----
From: Martin Sebor [mailto:sebor@roguewave.com] 
Sent: Tuesday, March 21, 2006 05:27
To: stdcxx-dev@incubator.apache.org
Subject: Re: test for 21.string.copy


Anton Pevtsov wrote:
> The attached file contains the test for lib.string.copy.

I made a few changes -- see the attached diff.

exp_exceptions[] was missing element[2] (dereferenced later in the
test). I added it.

I introduced the LSTR and LLEN convenience macros to make it easier to
refer to long_string and long_string_len.

I made TestCase::str and TestCase::res const to avoid having to cast
away the constness of string literals used to define the array and
silence gcc warnings about the same.

I made the MemFun pointer argument const since pointed-to object is not
modified by any of the functions. I replaced the s_empty object with a
constant eos character for simplicity. I also added const wherever
possible. Making all objects that aren't modified makes code easier to
understand. Let's use const wherever possible.

I added a few assertions to assert test preconditions. Please use the
RW_ASSERT() to assert any assumptions made in all tests :)

I initialized the destination buffer to all garbage. Could you make sure
the test verifies that the tested function doesn't write past the end of
the provided buffer (by checking for the garbage value)?

Finally, I removed the extraneous "test disabled" text from the
rw_info() statements (probably copy-and-paste mistakes :)

With the patch applied and the suggested changes to test for the
potential corruption of the destination buffer, please go ahead and
commit the test.

Thanks
Martin

PS I added rw_narrow() overloads that you could use to display the
contents of the (wide) destination buffer until we have the formatting
directive implemented (I'm working on it).

Also, I noticed some strange long --trace output in a few cases that
doesn't look right to me (see below). Could you look into it to see
what's going on? (Is it because we're not using the right formatting
directive?)

# ITRACE_ASSERTION (S0) (4 lines):
# TEXT: line 215. std::basic_string<wchar_t, char_traits<wchar_t>, 
allocator<wch
ar_t>>({ 'x' <repeats 4095 times> }).copy ("\0", 4095) expected 
"\0\0\0x\0\0\0x\
0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\
0\0x\0\0
...
\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0", got 
"\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\
...
0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\0x\0\0\
0x\0\0\0
x\0\0\0"
# CLAUSE: lib.string.copy
# LINE: 133