* fix test 26481: it tests for something different than the incorrect type parameters and in fact the need to fix this highlights the need for the previous changes

* additional tests added similarly to tfexpand.pp to test modified behaviour discussed in bug 0034166

* Modified test to not create sub-dirs outside the test's directory.

* Define slowcpu for android.

* Erase the temp file.

+ lazarus project files for the avx assembler test generators

* fix to write correct output file name

+ support fpc include output in avx test generation code

* fix tests now that the RTTI of method pointer variables also contains the hidden parameters

* Delete temp files after test.

* fix tests as depending on the platform the first parameter in the RTTI data might *not* be $self

* tests: On x86_64-android long double is 128-bit. There is no support for 128-bit floats in FPC yet. So disable long double tests for x86_64-android for now.

* Disabled some tests for android since they turn PIC off.

* Disabled some tests for android.

+ added an interactive keyboard unit test

+ slight fixes and improvements to the test

* fix test for windows

* print ASCII characters better

* write physical key numbers both in decimal and hex

* show all details about the key event, without using KeyEventToString (because it
sometimes outputs control (<#32) ASCII characters)

+ added a keyboard dump test tool

* number of cycles slightly increased for better measurements

* moved some drawing routines to a new unit 'vidutil'

+ added an interactive keyboard test

+ added keyboard dumps, obtained with FPC 3.0.4 under Windows, using the US keyboard locale

+ added kbd dump file with num lock on

+ added a keyboard dump using the Bulgarian Phonetic Traditional keyboard layout and code page 866

+ added a keyboard BIOS dump tool

+ added BIOS int 16h keycode dumps

+ added a LibreOffice spreadsheet with the BIOS key codes, as dumped in the .bdf files

- removed the executable flag from the *.bdf files

+ added bios key dump with the left alt and left shift combination pressed

* the 'left ctrl + left shift' column moved to the right of 'right ctrl'

* .bdf files renamed .bios.txt

+ added the lalt+lshift bios keycode dump to the spreadsheet

* right justify text columns in spreadsheet

+ added color codes for the keys

* address output updated

* made test more verbose

+ added the 102-th key on international keyboards (outputs "<" or ">" with the US layout)

+ added test, that reads enhanced keyboard events

* use volatile() to prevent wrong optimizations by llvm

+ added all the missing test/units/* subdirectories: classes character dateutil fpcunit softfpu variants

+ test for previous commit

* updated bcase.pp benchmark by Gareth Moreton, now used also as test

+ test for 0027543 which is fixed meanwhile

haiku-x86_64: regenerated all Makefiles

tarray15: no rodata support on a bunch of m68k platforms

tarray15: the conditions added in 0040779 should be or, instead of and (thanks Pierre)

* usage example for linux

+ support for 16 bit registers

+ added a simple test that tests TList.Sort

Add testing of use of variable of class

* casing fixed

Commit patch submitted by J. Gareth Moreton in bug report 35180

* only enable tw35187 for Windows and Linux, since it uses syscalls under unix
(which are not supported on all unix targets; feel free to re-enable for
specific unix targets if the test has been verified to compile there)
o also fixed placement of dotest directives: they have to come before any
part of the program code

+ test from mantis 0035224 (already works)

Avoid arithmetic overflow inside heaptrc unit on this test

Fix compilation of testsuite program

Link against libc to avoid __guard_local not found symbol failure at link time for OpenBSD target

Link against libstdc++ and libc to avoid linking error for openbsd OS

- removed explicit linking of ptest.o, as that happens already automatically
due to the "uses ptest" and breaks LTO (then both ptest.o and ptest.bc
get linked, resulting in duplicate symbols)

- removed test that shortstring is aligned to a multiple of pointer size,
as the compiler does not guarantee any such thing

* disabled low-level hack test for LLVM

* replaced non-breaking space with space (patch by J. Gareth Moreton,
mantis 0035602)

+ Added test object files for x86_64-haiku.

* use variables of correct size with movd (test checks whether movd only
reads 32 bit even when destination is xmm register) (mantis 0035701,
as noted by J. Gareth Moreton)

* forgot to commit tests in 0042329

* rename custom attributes related tests to something more sensible

* also check for invalid attribute data in tests

* fix test

+ add a test for overloaded attribute constructors

* fix test on systems where Currency is calculated using floating point types

* we halt if the currency values are *not* equal, not if they're equal... m(

* fix Currency comparison... again...

- removed tests as modern Delphi versions behave differently (mantis 0035886)

* fix compilation of tests on 32-bit systems (doesn't (shouldn't?) change their behavior)

* correctly align the stream on targets that require proper alignment

* compilation of tests on 32 bit targets fixed, using while loops so keeping the type sizes

* test needs exception checking

* fix test for arm

* test from mantis 0035897, already fixed

* change result of pi function from ansistring into dynamic array, so we
generate unique calls every time we write to it
* replace calls to high(ansistring/dynarray) in the inner body of the pi
function with a variable that gets assigned once
-> the FPC version is now just as fast as the C version

+ add a test that ensures that the introduction of fullscale array constructors didn't mess with array constructors passed to an open array of Variant

* added error checking to the test

Move test for read-only data to new separate test source treadonlydata.pp

* Revert changes to tarray15.pp
* Modify treadonlydata.pp to not use dynamic array.
+ Add test treadonlydata2.pp that should fail at compile time.

* initialise local variable to prevent random failures

* make halt codes unique

* extend test so that it also covers the SetToString/StringToSet variants take/return Integer arguments

+ extra test to test remap functionality with -gh

* test properly (with fno-exceptions) compiled

* test fixed for i386

* fix memory leak of tcustomattr13

* fptime works now usefully on unix systems
+ support -n to run the command times
+ display average runtime when -n is passed

* test for tw36212

* Testsuite in punit format

* Remove testsuite again

+ readded accidently deleted file

* naming of tests fixed

+ add test for a problem reported by Michael van Canneyt which is fixed by the previous revision

* patch by J. Gareth Moreton: improve test and benchmark, resolves 0035633

+ test for 0026791, meanwhile fixed

Avoid using <> for floating point values, check that abs of difference is small instead (fixes m68k 'pseudo-failure')

* tests: When cross-testing, if possible, use the "ppc" native compiler to build fpmake. It solves PPU incompatibility errors.

* Skip the test for aarch64.

* Skip the test for android.

* test fixed

- test removed as it cannot work on all variants of currency support

* Disabled range and overflow checks for the tw36587 test, since overflows are intended there.

* tests need -CE

* aarch64 suffers also from the limited tan precision of the softfpu

* AArch64: the PIC code needs also more registers, so use the same ifdefs as arm and sparc

+ add test for already fixed Mantis 0024283

+ with the randseed set as in the test, the test fails on x86_64-linux, added so it won't be forgotten

* commented test

* fixed test for 32 Bit targets

* cast only the longint

* fixed tests
o fixed use of loop counter after end of loop
o fixed reliance on old behaviour for int -> real typecasts

* add tests\test\unit\types directory to list of tested directories

* enable tests for the Nullable unit

* fix test to check result properly as pointed out by Bi0T1N in the bug tracker

Fix ShowHistory, by avoiding ordering by TR_ID which is not monotoneous anymore

+ test for last commit

+ add test for 0031714 which is already fixed

* added test for Mantis 0036014 which is already fixed (also in 3.2.0)

* added test for bug that's already fixed (mantis 0037423)

+ forgotten test

* it's usefull to enable the test for all platforms

+ add test for already fixed Mantis 0037286

* Added the $F+ switch (forced far calls) for some TP mode tests.

* Fixed tests for 16-bit CPUs.

* made test more usefull

* non-zero exit code on failure

* fix test for targets with little memory

* made test more verbose

+ created xtensa-freertos object files using gcc xtensa-esp32-elf-gcc (crosstool-NG esp-2019r2) 8.2.0 with:
make TEST_CCOMPILER=xtensa-esp32-elf-gcc create_c_objects TEST_FPC=fpc TEST_ABI=windowed TEST_OS_TARGET=freertos TEST_CPU_TARGET=xtensa TEST_SUBARCH=lx6 TEST_CFLAGS=-mlongcalls

+ common include file for tcal*.pp tests to define cpusmall

* unified settings for CPU "size"

+ forgotten file

* replace tcaldefs.inc by cpudefs.inc

* regenerated

+ C objects for for xtensa-freertos with call0 abi, build with:

make create_c_objects TEST_CCOMPILER=xtensa-lx106-elf-gcc TEST_ABI=call0 TEST_FPC=fpc TEST_CPU_TARGET=xtensa TEST_OS_TARGET=freertos "TEST_CFLAGS=-O3 -mlongcalls"

using xtensa-lx106-elf-gcc (crosstool-NG crosstool-ng-1.22.0-100-ge567ec7) 5.2.0

* added missing "end." (fixup for 0046904)

* test for mantis 0034543 (already fixed)

* test for new split option

* test needs delphi mode apparently

+ Added a test for: When {$mode delphi} and {$modeswitch nestedprocvars} it is allowed to assign a nested routine which does not use parentfp to a regular procvar. And then call this procvar without any side effects.

+ test for 0038022

* add unix timezone test

+ test for mantis 0036381 (seems already fixed)

* fix copy-paste mistake from Mantis (due to replacement of '#' with links...)

* do not run the test, it is only about a compilation issue

add new testmethods (memref-operands)

add new testmethods (memref-operands)

working on new testmethods (memref-operands)

working on new testmethods (memref-operands)

working on new testmethods (memref-operands)

working on new testmethods (memref-operands)

working on new testmethods (memref-operands)

working on new testmethods (memref-operands)

* fix test for targets not having a real wide string

* fix test for x86_64 targets using the SysV abi

* meanwhile all linux targets support the checkpointer directive

* test needs -Cg to trigger the warning

* test for min/max intrinsics

bugfixing operandsize from 32-bit to (correct) 64-bit

bugfixing operandsize from 32-bit to (correct) 64-bit

+ Added a test for proper setting of fc_no_direct_exit in flowcontrol.

add new option to avxtestgenerator to list opcodes-memrefsize-state

add new option to avxtestgenerator to list opcodes-memrefsize-state

working on new testmethods and bug-fixing size of (memref-operands | global and local variable

working on new testmethods and bug-fixing size of (memref-operands | global and local variable

* extended test

* made exit codes unique

avx512: external gas assembler not supported gather/scatter-opcodes without mask-register =>> not generate testcode

avx512: external gas assembler not supported gather/scatter-opcodes without mask-register =>> not generate testcode

avxtestgenerator: add new commandline param '-m', create only these testfile(s)

avxtestgenerator: add new commandline param '-m', create only these testfile(s)

* simplified test after 0047892

Check also that SetRoundMode returns correctly previous rounding mode

Add testing for single float type in trndmode.pp

Add two new button to simplify inspecting evolutiontestsuite difference in time

Add check for bug fix in bug report 38339

bugfix testcases 000037785

bugfix testcases 000037785

avxtestgenerator add attsuffix in 'opcode-memrefsize-state - list' (command-line option '-l')

avxtestgenerator add attsuffix in 'opcode-memrefsize-state - list' (command-line option '-l')

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

Change parameter type of Build* procedure from out to var, as the StringList is created in calling main
This avoid a warning during compliaztion and a crash when compiled with -gt* option.

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

Add some colors for comparison between two run results

Avoid range check errors
(cherry picked from commit 1351ccb3bfe6adad6be0246acb7645203df8b3b0)

# Conflicts:
# .gitattributes

Avoid range check errors

* let theapthread wait unneededly less so the total run time is shorter

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

+ add a test which ensures that a "const TVarData" parameter is passed as a reference.
This is required for Delphi compatibility as implementers of IVarInvokable or
inheritors of TInvokableVariantType need to modify the variant data by using
a pointer to the TVarData because it's passed as const and thus not modifyable
by itself.
This behavior is documented in so far as the C++ builder documentation shows
that the same parameter is implemented as "const&".

+ added simple test for the video unit

+ added unicode video unit test

+ added lazarus projects for the video unit tests for Windows

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new testmethods avx512-memref-operands - 'compressed disp8*N'

* test for already fixed issue 0032139

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

* test for already fixed issue 0024434

* .lpi updated

+ enabled all debug code generation options in the video unit test .lpi files

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

+ added simple test for the TRawByteStringCollection object type

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

Fix tw28713 for big endian CPUs

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

Extend list of CPUs defining slowcpu conditional in tmt1 and tw8177 sources

+ added extra unit paths to rtl-unicode

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

avxtestgenerator: working on new tests for compressed disp8*N

avxtestgenerator: working on new tests for compressed disp8*N

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

* fix test so it really uses random values for the boolean, thanks Jonas for the hint!

Update to objects compiled using GNU gcc 5.5.0 for sparc-solaris

+ new test

* fix test for RiscV

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

working on new testmethods avx512-memref-operands - 'compressed disp8*N'

* re-build c objects with a fixed C compiler regarding floating point parameter passing: riscv64-linux-gnu-gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0 with -O3

* test for bin2hex

+ compiled C object files, generated by riscv32-unknown-linux-gnu-gcc (GCC) 10.2.0 with -O3

fpts2junit: remove ../ at start for the classname

* test needs objfpc mode else i is 16 bit only

* added test, issue is fixed by 0048758 already

* { ... } is no a longer valid comment in aarch64 assembler

+ patch by J. Gareth Moreton: improved tests for constant divisions including benchmark,
second part of 0038806

+ added a test for Get/SetTextAutoFlush feature

* fix for wrong nested comments resulting in wrong results under *nix, etc., with redirected output

* fix for wrongly chosen conditional define name

+ test for already fixed 0026016

+ add a test for Variant dispatch and the generated call description format (this is for the previous three fixes)

- remove unused copy of tw17904 (wrong extension)

* Test from werner pamler for issue 0038968

* fix test so that it will work correctly on non-Windows as well

Fix compilation of tiso8601 test by adding -Mobjfpc option

* set the exeext for WASI properly in dotest.pp

* Test for fix of issue 0038489

* Restored disappeared test

fpts2junit: prevent double entries in junit xml (in test log are double entries, one for compiles, one for runs)

+ Added a test for the TP-style INLINE() statement.

Add support for hash generation for git repository (local modifier not implemented)

fcl-json: fixed testjson.lpi

Add test for bug report 38973 (already fixed in trunk/main)

* tw39323 needs to turn on range check to test what it is supposed to test

* Explicitly specify CP1252 for registry tests which depends on this CP.

* Weak externals from DLL are nor supported on Windows.

* Disabled running of the sharemem test since fpcmemdll.dll is never built.

* Prevent a crash of the tw2423 test on some versions of Windows.

* Modified the test to run properly on non-UTF8 systems.

* Explicitly specify DefaultRTLFileSystemCodePage for the test in order
to be independent from the current system code page.

* Do not run the tzlib1 test. It is not intended for this.

* fix test

Update to use gitlab

* bug fix in test taddbool.pp - it was testing longbool conversion to boolean
twice, and not testing wordbool conversion to boolean at all

+ added WASI .o files for the tests for linking with C

+ added test for SysUtils.GetLocalTime

* test for gitlab 0037785 which is resolved meanwhile

Added missing inc file

* pass a randseed which is known to be problematic

* extended test to cover last commit

* Add test for case of pointer and size

* consecutive naming

Add some delay and a maximum attempt number in GenerateTempDir to avoid getting oversized execution log files

* Restore test project

Added two new overflow tests

Converted tabs to spaces in toverflow.inc for consistency

+ add a test for converting a 64-bit Boolean (either QWordBool or Boolean64) to a 32-bit signed integer (this problematic case was discovered after e65b822a on PowerPC

* files moved as strutils is now in rtl-objpas

* activate missing test

New POPCNT tests to evaluate new optimisations

Fix compilation of the two tests for jvm-android target

* fix test for improved fpc_Val_SInt_ShortStr

+ add test for already fixed gitlab 0022252

+ test

* enable brenchmark to be compiled with -Sh

* tests fixed after last commit

* test does not need to use sysutils anymore

* clean up

* do not use sysutils anymore

+ test for already fixed bug so this avoids that is is reintroduced

Add possibility to pass jvm target compilation parameters and clean up if no problem is detected

+ added wasi to the list of targets without thread support

Added test to check WideChar const to UTF8String const conversion

* Use correct define for tests 51/52

Retry incase of thread initialization failure

fcl-passrc: clean up

* Restored test unit

+ add tests for function references, anonymous functions and C-blocks

pastojs: comment

pastojs: started test TPas2JSUseAnalyzer

* Test for const list

+ add C/C++ object files for aarch64-win64 compiled with LLVM 13.0.0

testobj3: fix test

Use RTL StringRefCount instead of local GetRefCount that only supports some
platforms and depends on the internal layout of ansistring headers

pastojs: less hints

+ test for last commit (44094dba2f96fcc6c252f9c42398b6740edb85a1)

+ support for generic forward declarations fixes gitlab 0039581, so add a test for it

+ support for generic forward declarations fixes gitlab 0039582, so add a test for it

* test for already resolved issue gitlab 0038874

Added test cases for AVR shift by compile time constant.

+ added second screen to the video4 test, which demonstrates all the attributes (without combining them)

fcl-passrc: test forbidden anonymous function types

* Test from Inoussa to test NormalizeNFD

* Test for stringtoidentifier

+ slowcpu define

+ test for gitlab 0039627, together with the last merge request it resolves gitlab 0039627

Add several tests for 127 max ID length, related to bug report gitlab 0039661

* New tests for aligned records

+ add test for already fixed gitlab 0039677

* more precise test as suggested here: https://gitlab.com/freepascal.org/fpc/source/-/commit/876ae561585c22b9809764710f11b9a88da935ee#note_921869640

* make test compiling on avr

* write number of iterations

* do not run test, it is about successful compilation and runtime on weak CPUs is high

db testsuite: add TestMSSQLLargeStrings

* Change writeln+exit to ignore

* for now, IO is not supported on AVR

* made test compilable if no floating point type support is available

* made checking for supported targets more fool proof

* New test for gitlab 0039713

Text for merge request #226

Fix %cpu type as spotted by Alexey Torgashin

fcl-json: fixed tests

* fix regressions after previous patch by correctly parsing output, which does not contain '##WASI-EXITCODE:'

* add cblock directive so that the test fails for the correct reason

Remove %OS by %TARGET, as spotted by Gareth

* test also cmem unit intensively

* tests fixed

+ add a mode Delphi equivalent for tanonfunc56 to check whether overload selection works correctly

+ add tests that ensure that the overload selection works correctly when functions, methods or nested functions are assigned when overloads for function references are involved as well

+ add tests to ensure that the assignment of nested function variables to function references is forbidden

fcl-passrc: transition TPasElement from reference count to free

* Improved TESTALL.BAT file for JVM tests

* experimental test for gitlab 0039758

* naming fixed, it is a test for gitlab 0039785

+ test for gitlab 0039786

* as correctly pointed out by Bart B, -gw instead of -g should be used in the test for gitlab 0039816

+ test for gitlab 0033559

* tests for merge request 244

webidl: fixed tests

* New test to catch i39851

+ added test for wasm globals

+ added wasm dir to test dirs

+ added test for wasmglobals, where the globals are declared in the implementation section of a unit

* test for gitlab 0039834, previous commits resolve gitlab 0039834

* experimental test for gitlab 0039722, no sure if it will work this way on all platforms

* extended test

* improved test as proposed by Rika