ARM IHI0031F "Arm Debug Interface Architecture Specification"
chapter C2.6.1 "BASE, Debug Base Address register" reports:
A debugger must handle the following situations as
non-fatal errors:
- ...
- An entry in the ROM Table points to a faulting location.
- ...
Typically, a debugger issues a warning if it encounters
one of these situations. However, Arm recommends that it
continues operating. An example of an implementation that
might cause errors of this type is a system with static
base address or ROM Table entries that enable entire
subsystems to be disabled, for example by a tie-off input,
packaging choice, fuse, or similar.
Don't halt ROM table parsing if one entry causes an error; log the
error condition and continue to next entry.
Not sure if we have to send an ABORT before continuing.
Change-Id: I94fdb5b175bfb07dde378149421582b7e7cd5b09
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6818
Tested-by: jenkins
Reviewed-by: Daniel Goehring <dgoehrin@os.amperecomputing.com>
As Class 0x1 ROM table, also Class 0x9 ROM tables encodes a flag
for system memory access.
Detect the flag in rtp_cs_component() and dump the same message
for both type of ROM tables.
Extend rtp_read_cs_regs() to read ARM_CS_C9_DEVID.
Change-Id: Ic85d1ea068ed706ceedfd65076ff4c96d04e9792
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6817
Tested-by: jenkins
Reviewed-by: Daniel Goehring <dgoehrin@os.amperecomputing.com>
During ROM table parsing, each ROM table entry points to a
CoreSight component that can, in turn, be another ROM table.
Split the specific code for ROM table handling from the generic
CoreSight code.
Log an error if a ROM table entry cannot be read.
Change-Id: I5ad106a99b9c21ddb48b5b162ae87101e4f49878
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6816
Tested-by: jenkins
Rework dap_read_part_id() while preparing for reorganizing the
'ROM Table Parsing' (RTP):
- rename it with 'rtp' prefix;
- extends it to read other CoreSight registers, thus improving the
overall speed by queuing more reads;
- reduce the list of arguments by using a struct;
- reorder the reads by increasing offset, potentially gaining
speed using MEM_AP_REG_BDx and/or auto-increment;
- log a debug message in case of read error.
Change-Id: I6544ac7740b808a6c0fbacf97ac00b97f5bd3832
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6815
Tested-by: jenkins
ADI v5.1 and v6.0 permit the definition of CoreSight components (class 9
ROM entries).
dap_rom_display() is refactored a bit such that we always end up with
attempting to parse the ROM contents using the appropriate upper limit
for class 1 and 9 ROM types.
Change-Id: I4ba497b3807f1f11f06186eb6e61959ea3540c59
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6359
Tested-by: jenkins
Use the list of values from ARM IHI0029E to decode and print the
Device Architecture register.
Add attribute 'unused' to the function, not used yet.
Change-Id: I7b1dd204bd1db671578c588372b667e23611876c
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6463
Tested-by: jenkins
Reviewed-by: Daniel Goehring <dgoehrin@os.amperecomputing.com>
In case of AP not responding, e.g. not clocked, the first WAIT
reply is logged as:
DAP transaction stalled (WAIT) - slowing down
then OpenOCD retries the transaction few times, until it timeouts.
At each retry it prints the message:
DAP transaction stalled during replay (WAIT) - resending
Depending on JTAG speed and transport latency, the amount of log
messages can be quite annoying and not relevant.
The last printed line is at timeout:
Timeout during WAIT recovery
Reduce the verbosity.
Change-Id: I5a7a337527c98b2450de59066b13713511c2894f
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6814
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-by: Daniel Goehring <dgoehrin@os.amperecomputing.com>
Right now it has a single use but it will soon be used more.
Change-Id: I9a819c65df467fc859e4b5251035a17ed33daa35
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6813
Tested-by: jenkins
Reviewed-by: Daniel Goehring <dgoehrin@os.amperecomputing.com>
since commit 12d1ad0c75 : update capstone include path ...
the generated capstone.pc is not working
so fix the includedir in capstone.pc to get github action working
Change-Id: I7767e181a74c73a7514eeb6293cd556a794dbfe9
Signed-off-by: Tarek BOCHKATI <tarek.bouchkati@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6969
Tested-by: jenkins
Reviewed-by: Tim Newsome <tim@sifive.com>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Append lacking e to word Note
Signed-off-by: Piotr Kasprzyk <ciri@ciri.pl>
Change-Id: Ibd40a2f93d11cf1945361f0c46329b88963d6826
Reviewed-on: https://review.openocd.org/c/openocd/+/6963
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
In jimtcl 0.80 the prototype of Jim_DictPairs() has changed.
The only code in OpenOCD that uses Jim_DictPairs() has been merged
recently and it only uses the current jimtcl syntax.
To allow compiling OpenOCD master branch with older versions of
jimtcl, detect the version of jimtcl and use the appropriate
syntax.
Change-Id: I6fc78303b6a4db064a97f326c46119f4568e88f3
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reported-by: dullfire@yahoo.com
Reviewed-on: https://review.openocd.org/c/openocd/+/6948
Tested-by: jenkins
Added bluenrg-lps support
Added file for the board steval-idb012v1
Fixed size_info information using a mask
Changed the if condition in bluenrg-x.cfg to be valid only for bluenrg-1 and bluenrg-2
Signed-off-by: Salvatore Giorgio PECORINO <salvatore-giorgio.pecorino@st.com>
Change-Id: Ic0777ec0811ee6fac7d5e1d065c4629e47d84a1f
Reviewed-on: https://review.openocd.org/c/openocd/+/6928
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Commit 5ebb1bdea1 ("server/gdb: fix return of gdb remote monitor
command") replaces the call to command_run_line() with call to
Jim_EvalObj() but does not properly set the "context".
In multi-target environment, his can cause the erroneously
execution of the command on the wrong target.
Copy from the code in command_run_line() the proper setup before
executing Jim_EvalObj().
Change-Id: I56738c80779082ca146a06c01bc30e28bc835fd3
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reported-by: Bohdan Tymkiv <bohdan200@gmail.com>
Fixes: 5ebb1bdea1 ("server/gdb: fix return of gdb remote monitor command")
Reviewed-on: https://review.openocd.org/c/openocd/+/6966
Tested-by: jenkins
Reviewed-by: Bohdan Tymkiv <bohdan200@gmail.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-by: Tim Newsome <tim@sifive.com>
The deprecation was already in the documentation since v0.11.0
through commit 85ba2dc4c6 ("rtos/hwthread: add hardware-thread
pseudo rtos") but OpenOCD was not informing the user printing a
runtime message.
Remove the deprecated method from the documentation and print a
deprecated message at runtime.
There is no reliable way to print the same message in GDB console,
so we have to rely on user noticing it in the OpenOCD log.
Target is to remove the functionality after v0.12.0.
Change-Id: Idd2d9e3b6eccc92dcf0432c3c7de2f8a0fcabe9f
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6862
Tested-by: jenkins
on macos (homebrew base) `pkg-config --cflags capstone` output with
`-I/opt/homebrew/Cellar/capstone/4.0.2/include/capstone`
gcc not find headers on parent "include" path,
causes build error `fatal error: 'capstone/capstone.h' file not found`
it's ok to change to <capstone.h> for all platforms.
Signed-off-by: fatalc <cnfatal@gmail.com>
Change-Id: Ia2e2058024d4fc1a57a8b4ea847c664d74f67efb
Reviewed-on: https://review.openocd.org/c/openocd/+/6946
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Handle JIM_CONTINUE return value of adiv5_jim_mem_ap_spot_configure(),
otherwise OpenOCD silently quits when an unknown option is provided.
Change-Id: I9b1351c0911e74999d8dd1260ede9760088510d7
Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/6944
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The correct ordering is required to prevent two outputs connected
together.
Change-Id: I634a9ca7e0ccf337d1723011b8aee1f2d81efbcf
Signed-off-by: Steve Marple <stevemarple@googlemail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6937
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The SWDIO buffer requires a direction pin to select input or output
direction. Output is selected by a high logic level (matches
bcm2835gpio driver).
Change-Id: I240cb99a5dfea08121bb33d4b5e2108ce7597468
Signed-off-by: Steve Marple <stevemarple@googlemail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6936
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
According to GigaDevice user manuals the devices have OBRLD bit in FMC_CTL
register which is functionally compatible with OBL_LAUNCH @ FLASH_CR
of STM32 counterparts.
Change-Id: I84d231b38815fcb6452fd73b9153b269cce3b737
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6759
Tested-by: jenkins
Reviewed-by: Andrzej Sierżęga <asier70@gmail.com>
The flash is compatible with stm32f1x, reuse the driver.
Extend the size of work area to RAM size of the smallest device.
Stop watchdogs before flash programming.
Change-Id: I67a7654a6e196f9d4b2409edaa7990c53334437e
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6711
Tested-by: jenkins
Reviewed-by: Tim Newsome <tim@sifive.com>
The device has compatible flash macro with STM32F1 family, reuse
stm32f1x driver code.
Detect non-ARM target - for simplicy test target type name 'riscv'
and the address has 32 bits.
In case of RISC-V CPU use simple chunked write algo - async algo
cannot be used as the core implemented in this device doesn't
allow memory access while running.
Change-Id: Ie3886fbd8573652691f91a02335812a7300689f7
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6704
Tested-by: jenkins
Reviewed-by: Tim Newsome <tim@sifive.com>
Zephyr plan to remove openocd specific symbols in favour
of more generic one.
These generic symbols has been introduced in Zephyr 2.6.
Signed-off-by: Julien Massot <julien.massot@iot.bzh>
Change-Id: I89418c9c378fb8b8baa29763fc6f1b6e652dc7ef
Reviewed-on: https://review.openocd.org/c/openocd/+/6844
Tested-by: jenkins
Reviewed-by: Stephanos Ioannidis <root@stephanos.io>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Fixed the reading of p_flags in ELF64 segment headers - that field
is 64 bits wide.
Change-Id: I053ca57d36efb54b7c638484acd6c7a2fbcbd05a
Signed-off-by: Jan Matyas <matyas@codasip.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6927
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Current implementation for gdb remote monitor command uses the
command_run_line() to execute the command.
While command_run_line() has several advantages, it unfortunately
hides the error codes and outputs the result of the command
through LOG_USER(), which is not what gdb requires. See 'qRcmd' in
https://sourceware.org/gdb/onlinedocs/gdb/General-Query-Packets.html
Replace command_run_line() with Jim_EvalObj() and parse the output
to provide the proper result to gdb.
Can be tested by defining in OpenOCD:
proc a {} {return hello}
proc b {} {return -code 4}
proc c {} {return -code 4 "This is an error!"}
then by executing in gdb console:
monitor a
monitor b
monitor c
monitor foo
Change-Id: I1b85554d59221560e97861a499e16764e70c1172
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reported-by: Torbjorn Svensson <torbjorn.svensson@st.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6886
Tested-by: jenkins
The target nds32 and its companion adapter aice have not received
any real improvement since 2013.
It has been hard to keep them aligned during the evolution of
OpenOCD code, with no way for maintainers to really check if they
are still working.
No real documentation is present for them in OpenOCD.
The arch nds32 has been dropped from Linux kernel v5.18-rc1.
Deprecate both nds32 target and aice adapter with the target of
dropping them for v0.13.0.
Remove automatic build of aice, forcing user to select it.
Change-Id: Ib465d676246fa3b4e95c3d399ba9a5cf1f8b3baf
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6887
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Tested-by: jenkins
The current code locks the flash controller in case of error during
flash write only. An error in other flash operations may cause the
flash is left unlocked.
Implement locking also after error in erase, mass erase, options
write and erase.
Change-Id: I26c2ed7914e7847122306f29b777b9eefd1dc580
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6710
Tested-by: jenkins
Reviewed-by: Tarek BOCHKATI <tarek.bouchkati@gmail.com>
stm32x_wait_status_busy() has two side effects in case of flash programming error:
- reports error
- clears error bit in status register
Use stm32x_wait_status_busy() to report also flash error during target
algo flash write.
While on it use more descriptive error codes in stm32x_wait_status_busy().
Change-Id: I6e1cffc2aa5411b918a23ed62d5194910888a9d1
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6709
Tested-by: jenkins
Reviewed-by: Tarek BOCHKATI <tarek.bouchkati@gmail.com>
Use target_get_working_area_avail() instead of try-fail iteration.
Call destroy_reg_param() in a for cycle.
Change-Id: I1891d1ffdea99010c6ab66b9578400b9d7922e20
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6708
Tested-by: jenkins
Reviewed-by: Tarek BOCHKATI <tarek.bouchkati@gmail.com>
Mostly refactoring.
Rename original stm32x_write_block() to stm32x_write_block_async()
as it uses target async algo.
Introduce new stm32x_write_block() and move slow, host controlled
fallback flash write there.
The change allows stm32x_write_options() to use slow flash write fallback.
While on it rename variables where halfword count is stored.
Change-Id: I386ae15cf052b1490461ed8f7eea5b4403d466f7
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6706
Tested-by: jenkins
Reviewed-by: Tarek BOCHKATI <tarek.bouchkati@gmail.com>
When we debug a target that works as a USB device, halting
the target causes the USB communication with the USB host to
become unresponsive. The host will try to reconnect/reset/setup
the unresponsive device during which communication with other
devices on the same USB bus can get stalled for several seconds.
If the JTAG adapter is on the same bus, we need to make sure
openOCD will wait for packets at least as long as the host USB
stack. Otherwise the USB stack might deliver a valid packet, but
openOCD would ignore it due to the timeout. The xHCI spec uses 5
sec timeouts, so let's use that in openOCD with some margin.
Use this value in all libusb calls. HID API might have a libusb
backend and would probably be victim to the same bug, so it
should use this timeout, too.
Ticket: https://sourceforge.net/p/openocd/tickets/343/
Signed-off-by: Gabor Csapo <gaborcsapo@google.com>
Change-Id: Ia3dc1356e676fe550f57a4c72f7a24ba296b6af2
Reviewed-on: https://review.openocd.org/c/openocd/+/6882
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
- Print also the target name, not just the packet contents.
This is important when there are more GDB servers (more
debug-able targets) active in one OpenOCD session.
- Log also the received Ctrl-C requests coming from GDB
(one byte 0x3), ACKs ("+") and NACKs ("-").
- Do not print zero-length incoming packets (this occurred
when Ctrl-C packets were received).
- Removed a stray apostrophe "'" that got printed
in gdb_log_outgoing_packet()
Signed-off-by: Jan Matyas <matyas@codasip.com>
Change-Id: If68fe0a8aa635165d0bbe6fa0e48a4645a02da67
Reviewed-on: https://review.openocd.org/c/openocd/+/6879
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
To support 64 bits bit and masks
Replace local definition of BIT in rtos/chromium-ec
Change-Id: I1f268d6e8790f1b07bf798680b797878ce81064b
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6857
Tested-by: jenkins
Including config.h as first is required for every C file.
Add it to the C files that still miss it.
Change-Id: I1a210e7d3a854958a85a290b086ad8a9f5176425
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6856
Tested-by: jenkins
While the documentation reports that SWO pin frequency can be
omitted to let the adapter autodetect the value, the code wrongly
drops an error when pin frequency is not specified.
Don't require the pin frequency to be set at "enable", but verify
that the adapter has properly changes it to a valid value.
Change-Id: I3dfbe3256e8887ef4f03512769b06381cdc9db0d
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reported-by: Karl Palsson <karlp@tweak.net.au>
Fixes: 184724d14e ("arm_tpiu_swo: add support for independent TPIU and SWO")
Reviewed-on: https://review.openocd.org/c/openocd/+/6310
Tested-by: jenkins
Reviewed-by: zapb <dev@zapb.de>
Tested on an EFM32PG12 Starter Kit.
Change-Id: I2cbc36fe0d2ad2089bf8c1e7d2260daaae4ddbb4
Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/5353
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Commit dbbac5f11d ("semihosting: use open mode flags from GDB,
not from sys/stat.h") fixes the conversion of the mode flags from
ARM semihosting encoding for SEMIHOSTING_SYS_OPEN to GDB mapping
for open().
Doing this, it breaks the conversion to local host's OS mapping
for open().
Split the conversion array to one for GDB and one for local host.
The local host conversion array is taken directly from the old
code.
Change-Id: I385321ddd32c3ac5cf6da3f1ce9eff76b05dd527
Fixes: dbbac5f11d ("semihosting: use open mode flags from GDB, not from sys/stat.h")
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reported-by: Erhan Kurubas <erhan.kurubas@espressif.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6870
Tested-by: jenkins
Some configuration of GCC could default to -fno-inline, causing
the build to fail after commit fb43f1ff4e ("target: Rework 'set'
variable of break-/watchpoints").
Switch the new inline functions to 'static inline', as it's widely
used in the rest of the code.
Change-Id: I8bf31045a137bd34ed825f4b2a9338eb3c70046d
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Fixes: fb43f1ff4e ("target: Rework 'set' variable of break-/watchpoints")
Reviewed-on: https://review.openocd.org/c/openocd/+/6881
Tested-by: jenkins
The 'set' variable name suggests a boolean data type which determines
whether a breakpoint (or watchpoint) is active. However, it is also
used to store the number of the breakpoint.
This encoding leads to inconsistent value assignments: boolean and
integer values are mixed. Also, associated hardware comparator
numbers, which are usually numbered from 0, cannot be used directly.
An additional offset is required to store the comparator numbers.
In order to make the code more readable and the value assignment more
consistent, change the variable name to 'is_set', its data type to 'bool'
and introduce a dedicated variable for the break-/watchpoint
number.
In order to make the review easier, the data types of various related
variables (e.g. number of breakpoints) are not changed.
While at it, fix a few coding style issues.
Change-Id: I2193f5639247cce6b80580d4c1c6afee916aeb82
Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/6319
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Considering this use case: (using STM32 L5 or U5)
1- first probe : TZEN enabled, RDP level 0
flash_regs_base |= STM32L5_REGS_SEC_OFFSET => 0x50022000
2- the user promotes the RDP to level 0.5
3- the second probe, fails to read OPTR using secure flags_regs_base:
used OPTR address is 0x50022040
Step 3 fails because when RDP is level 0.5, we should use Non-Secure
flash registers.
To fix this, always use NS flash regs to read OPTR in probe functions.
Fixes: 80d323c6e8 (flash/stm32l4x: introduce auto-probe when OPTR is changed)
Change-Id: I296aa633972b0c410b927488c999584a07b912d3
Signed-off-by: Tarek BOCHKATI <tarek.bouchkati@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6864
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Tested-by: jenkins
ST-Link Server 2.1.0-1 fixes concurrency issue with RW_MISC command
Starting from this version the ST-Link Server API is now v3.
In this change we save the ST-Link Server version, and check if the
API is greater or equal to 3 to enable the queuing.
Change-Id: I239eb81024700514c607a269b66651f457206faa
Signed-off-by: Tarek BOCHKATI <tarek.bouchkati@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6876
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Antonio Borneo
Florian Fainelli
Tarek BOCHKATI
Piotr Kasprzyk
Tomas Vanek
Salvatore Giorgio PECORINO
jihongbin
fatalc
Marc Schink
Steve Marple
Julien Massot
Jan Matyas
Markus Reiter
Gabor Csapo