- Olena - LRE

10.35: use static offset.
by Nicolas Burrus 07 Nov '03

07 Nov '03

Ca marche plutôt bien, on l'enlèvera si ca pose pb ... Index: metalic/ChangeLog from Nicolas Burrus <burrus_n(a)lrde.epita.fr> * mlc/type.hh (any_with_diamond): Use the static offset technique. Index: metalic/mlc/type.hh --- metalic/mlc/type.hh Mon, 01 Sep 2003 20:12:38 +0200 burrus_n (oln/c/21_type.hh 1.26 640) +++ metalic/mlc/type.hh Fri, 07 Nov 2003 17:19:10 +0100 burrus_n (oln/c/21_type.hh 1.27 640) @@ -34,6 +34,8 @@ # include <string> # include <cassert> +// FIXME: make a hierarchy of any with the different implementations + // any and any_with_diamond types are not defined in namespace mlc to // avoid namespace clashes. Indeed, a type inheriting from another in // a namespace foo will have implicit access to the foo namespace. @@ -53,12 +55,10 @@ E& exact() { - assert(offset_assigned); return *(E*)((char*)this - exact_offset); } const E& exact() const { - assert(offset_assigned); return *(const E*)((const char*)this - exact_offset); } @@ -67,25 +67,30 @@ return std::string("any_with_diamond<") + E::name() + ">"; } - public: + private: // This stores the actual value of the offset between the this - // pointer and the this pointer of the exact type. This enables - // diamond hierarchies, as static_cast cannot perform it and - // reinterpret_cast or (Exact*)(void*) cast are unsafe and - // compiler dependent. These members should be initialized by - // every concrete classes, using the mlc_init_static_hierarchy - // macro. - static int exact_offset; - // This Boolean determines if the exact_offset value has already - // been affected. - static bool offset_assigned; + // pointer and the address of the exact type. This enables diamond + // shaped hierarchies (static_cast can't and reinterpret_cast or + // (Exact*)(void*) cast are unsafe and compiler dependent). + static const int exact_offset; + + // Utility members used to determine the exact_offset + static const E exact_obj; + static const any_with_diamond<E>& ref_exact_obj; }; template <class E> - int any_with_diamond<E>::exact_offset = 0; + const E any_with_diamond<E>::exact_obj = E(); + + template <class E> + const any_with_diamond<E>& + any_with_diamond<E>::ref_exact_obj = any_with_diamond<E>::exact_obj; template <class E> - bool any_with_diamond<E>::offset_assigned = false; + const int any_with_diamond<E>::exact_offset = + (const char*)(void*)(&any_with_diamond<E>::ref_exact_obj) + - (const char*)(void*)(&any_with_diamond<E>::exact_obj); + // Warning, this class does not allow diamond hierarchies, consider // any_with_diamond if you really need it. @@ -284,7 +289,14 @@ | mlc_init_static_hierarchy | `--------------------------*/ -# define mlc_init_static_hierarchy(Exact) \ -mlc::assign_exact_offset<Exact, mlc::final>::doit(this) +// # define mlc_init_static_hierarchy(Exact) +// mlc::assign_exact_offset<Exact, mlc::final>::doit(this) + +// FIXME: this is for backward compatibility purposes only. This +// macro is called in every constructor of the classes, so we let them +// in place for the moment until we are sure that the static offset +// method works fine. + +# define mlc_init_static_hierarchy(Exact) #endif // ! METALIC_TYPE_HH

1 0

10.34: Fix pgm headers
by Nicolas Burrus 03 Nov '03

03 Nov '03

Le patch de mandor est appliqué. Index: olena/ChangeLog from Nicolas Burrus <burrus_n(a)lrde.epita.fr> * oln/io/pnm_write_2d.hh: Store in the header the maximal value of the encoding, not of the actual image values. * tests/io/tests/headers: New test. * tests/io/Makefile.am: Remove headers.pgm. Index: olena/tests/io/Makefile.am --- olena/tests/io/Makefile.am Tue, 05 Aug 2003 21:39:33 +0200 burrus_n (oln/e/49_Makefile.a 1.23 640) +++ olena/tests/io/Makefile.am Mon, 03 Nov 2003 21:49:33 +0100 burrus_n (oln/e/49_Makefile.a 1.23 640) @@ -6,6 +6,7 @@ CLEANFILES += \ black-test.pbm white-test.pbm \ + headers.pgm \ lena \ lena.ppm lena.pgm lena.pbm \ lena.pppm lena.ppgm lena.ppbm \ Index: olena/oln/io/pnm_write_2d.hh --- olena/oln/io/pnm_write_2d.hh Sat, 27 Sep 2003 18:30:39 +0200 burrus_n (oln/u/44_pnm_write_ 1.4 600) +++ olena/oln/io/pnm_write_2d.hh Mon, 03 Nov 2003 21:55:41 +0100 burrus_n (oln/u/44_pnm_write_ 1.4 600) @@ -129,9 +129,7 @@ pnm2d_info info; info.cols = im.ncols(); info.rows = im.nrows(); - oln::utils::f_minmax<oln_value_type(I)> f; - traverse(f, im); - info.max_val = f.max(); + info.max_val = ntg_max_val(oln_value_type(I)); if (info.max_val > ntg::to_ntg(65535U)) return false; Index: olena/tests/io/tests/headers --- olena/tests/io/tests/headers Mon, 03 Nov 2003 21:56:54 +0100 burrus_n () +++ olena/tests/io/tests/headers Mon, 03 Nov 2003 21:49:33 +0100 burrus_n (oln/v/32_headers 644) @@ -0,0 +1,38 @@ +// -*- c++ -*- + +#include "loadsave.hh" + +#include <fstream> +#include <sstream> + +template <class T> +bool check_type() +{ + image2d<T> ima (5, 5); + io::save(ima, "headers.pgm"); + + // Read the actual maximal value in the header of the written file + std::fstream f ("headers.pgm"); + assert(f); + std::string line; + for (unsigned i = 0; i < 4; ++i) + std::getline(f, line); + f.close(); + + // It should be the same value as ntg_max_val(T) + std::istringstream s (line); + unsigned value; + s >> value; + return value != ntg_max_val(T); +} + +bool +check(void) +{ + bool fail = false; + + fail |= check_type<int_u8>(); + fail |= check_type<int_u16>(); + + return fail; +}

1 0

oln 10.33: fix inheritance of not_integre_datatype images
by Nicolas Burrus 02 Nov '03

02 Nov '03

Index: olena/ChangeLog from Nicolas Burrus <burrus_n(a)lrde.epita.fr> * oln/core/abstract/image_with_type_with_dim.hh (image_with_type_with_dim_swith): Select data_type_image_with_dim by default, not data_type_image. * tests/core/tests/without_ntg: New test. Index: olena/oln/core/abstract/image_with_type_with_dim.hh --- olena/oln/core/abstract/image_with_type_with_dim.hh Thu, 07 Aug 2003 02:37:23 +0200 burrus_n (oln/v/11_image_with 1.3 600) +++ olena/oln/core/abstract/image_with_type_with_dim.hh Sun, 02 Nov 2003 17:27:23 +0100 burrus_n (oln/v/11_image_with 1.4 600) @@ -182,7 +182,7 @@ non_vectorial_image_with_dim<Dim, Exact>, mlc::bool_case_<true, - data_type_image<Exact> > + data_type_image_with_dim<Dim, Exact> > > > > > > >::ret ret; }; Index: olena/tests/core/tests/without_ntg --- olena/tests/core/tests/without_ntg Sun, 02 Nov 2003 17:27:37 +0100 burrus_n () +++ olena/tests/core/tests/without_ntg Sun, 02 Nov 2003 17:27:23 +0100 burrus_n (oln/v/31_without_nt 1.1 600) @@ -0,0 +1,16 @@ +// -*- c++ -*- + +#include <oln/basics2d.hh> +#include <iostream> + +class foo +{}; + +int main() +{ + oln::image2d< foo > i(60, 60); + + std::cout << i.hold(0, 0) << std::endl; + + return 0; +}

1 0

The 30 Hour Wonder! 627136814
by 240234＠aol.com 31 Oct '03

31 Oct '03

Cyalus comes in smaller doses and produces fewer side effects. Cyalus also works much faster than Sildenafil-based medications. In clinical trials, the majority of men who took the drug were able to engage in sexual intercourse within 30 minutes or less. The studies also indicated that Cyalus stays in the system for up to 30 hours. That's 26 hours longer than the traditional Sildenafil-based medications treatment. Test results showed that out of 700 participants at least 88 percent of men experienced improvement with their erections. More info http://www.hostvy.com/ 5B70BEAF-17F91E61-39C4105-29F5667B-25000864 6087EDA5-1C4C3A02-34F2E88A-619D8CF-4802E5AF 600DA917-39EBC767-6E88DB94-30C787D4-1EEF3B84

1 0