Re: [pdf-devel] Error report, few patches and part of testcases for pdf_time module.

[Aleksander Morgado]

Hi Łukasz,

Good to know that errors started to show up in the time module!

Jose, I would push those changes to the repo, and once done start fixing
time module, not before.

Cheers,
-Aleksander

Łukasz Spintzyk wrote:
> Errors gnuPDF
> 
> 1. pdf_time_init() -->pdf_time_context_init()
>     pdf_dealloc(time_struct) cause an error.
>     From man localtime
>         "The return value points to a statically allocated struct which
> might be overwritten by subsequent calls to any of the date and time
> functions."
>     so line
> pdf_dealloc(time_struct);
> is not necessary.
> Patch fixing this error(by deleting this line) -->pdf_time_init.patch
> 
> 2. pdf_i64_div
>     Division 1 by 86400 cause floating point exception
>     I have added another testcase to base/types/pdf-i64-div.c
>     which is in patch file pdf_i64_div_newTest.patch
>     This is that testcase
>         /*
>         * Test: pdf_i64_div_005
>         * Description:
>         *   Checks if the division 1/86400 gives 0.
>         * Success conditions:
>         * The call should not produce an error
>         */
>         START_TEST(pdf_i64_div_005)
>         {
>         pdf_status_t p_status = PDF_OK;
>         pdf_i64_t div1,div2,dest;
>         div1 = pdf_i64_new(0,1);
>         div2 = pdf_i64_new(0,24*3600);
>        
>        
>        
>         #ifndef PDF_USE_BUILTIN_64BIT_SUPPORT 
>         pdf_i64_div(&dest,div1,div2, &p_status);
>         fail_if(p_status != PDF_OK);
>         fail_if(dest.low != 0);
>         fail_if(dest.high != 0);
>         #else
>         pdf_i64_div(&dest,div1,div2, &p_status);
>         fail_if(dest != 0);
>         #endif
>        
>         }
>         END_TEST   
> patch name with that testcase is pdf_i64_div_newTest.patch
> 
> 
> 3. Functions :
>     pdf_status_t pdf_time_add_span()
>     pdf_status_t pdf_time_sub_span()
>     pdf_status_t pdf_time_diff()
>     and other functions in file src/base/pdf-time.c doesn't propagete
> status variable, but always return PDF_OK.
>     Some of them don't check if pointer argument is null.
> 
> 4. Patch which adds 22 of 38 files with testcases of pdf_time module to
> Makefile.am and runtests.c .
>     Patch name is pdf_time_22of32_tests.patch
>     To this mail are also appended archive time.tar.gz with that
> testcases, which schould be put in
>     trunk/torture/unit/base/time/
> 
> 5. pdf_time_add_cal_span
>     Both test from base/time/pdf_time_add_cal_span fails because of this
> same error.
>      In test case 002 there is:
>     /* test objects */
>         pdf_time_t time27, time28,time29, time30;
>           struct pdf_time_cal_span_s calspan;
>     /* initializing
>         ....
>     */
>         calspan.sign = PDF_FALSE;
>         calspan.years=0;
>         calspan.months=1;
>         calspan.hours=0;
>         calspan.minutes=0;
>         calspan.seconds=0;
>        
>         calspan.days=27;
>         pdf_time_add_cal_span(time27, &calspan);
>         calspan.days=28;
>         pdf_time_add_cal_span(time28, &calspan);
>         calspan.days=29;
>         pdf_time_add_cal_span(time29, &calspan);
>         calspan.days=30;
>         pdf_time_add_cal_span(time30, &calspan);
>    
>      fail_if(pdf_time_cmp(time27, time28) == 0);
>       fail_if(pdf_time_cmp(time28, time29) == 0);
>       fail_if(pdf_time_cmp(time29, time30) == 0);
>    
>     All of pdf_time_t objects have the same time, where only time27
> schould be valid and point to 1970-02-28 00:00:00. time28 schould point
> to 1970-03-01 00:00:00 and so on.
> 
> 6. pdf_time_from_string
>     6 of 13 test fails. This are mainly easy to fix bugs. I had put some
> comments in tests that fails.
> 
> 
> Thats is all rest of test schould be added in 2-3 weeks. However fixing
> bug in pdf_time_add_cal_span would be handful since it can broke other
> tests.
> BTW if you agree with bugs from point 3 and 6 i can write a patch for
> that functions too.
> 
> thanks
> Luke
>