Upgrade to GCC 10.1 toolchain #6294

earlephilhower · 2019-07-13T00:33:34Z

All examples I've tested run fine. GDB is working. No warnings.

earlephilhower · 2019-07-13T00:44:52Z

Lots of parenthesis warning, but basic examples like "Blink.ino" and "BearSSL_Validation.ino" run fine as far as I can see now. Still optimization on exceptions and removing the warnings which are causing the plain Arduino CI builds to fail.

Remove dependencies on earlier forked GNU utilities (gcc-xtensa, binutils-gdb-xtensa) and just use GCC sources, unmodified (except for patches in the esp-quick-toolchain directories).

GDB works with pure GNU GCC and pure GNU binutils now. Still warnings galore, but tested with the example sketch in the docs.

@d-a-v

Fix some warnings present in GCC8/9 in the IPAddress code In AddressListIterator there was a copy constructor which simply copied the structure bit-for-bit. That's the default operation, so remove it to avoid the warning there. IPAddress, add a default copy constructor since the other copy constructors are simply parsing from one format into a native ip_addr_t. @d-a-v, can you give these a look over and see if they're good (since IP stuff is really your domain).

These both use shared-ptrs to handle refcnts to allocated data, so using the default copy constructor is fine (and has been in use for a long time).

Make GCC think _heap_start is large enough to avoid the basic (and incorrect) bounds-checking warnings it produces. The size chosen is arbitrary and does not affect the actual size of the heap in any way.

earlephilhower · 2019-07-13T17:44:02Z

@d-a-v, could you check the AddrList and IPAddr changes?

d-a-v · 2019-07-13T22:33:24Z

Comparing a sketch with .text1 fitting issue, it appears file _divdi3.o takes much space in iram

      0x3df /path/to/libgcc.a(_divdi3.o)

This symbol is not here with gcc4.8.
Apart from that, I also run BearSSLValidation and it works great here too.
That's a great leap for espkind !

Instead of a bogus size, use an indefinite size for the heap to avoid GCC warnings

earlephilhower · 2019-07-14T02:54:58Z

Thanks, @d-a-v. I haven't been looking at the iram yet, still trying to get rid of the warnings. Are you OK with the changes to IP classes here? 3afe9e3

-edit- Just added the divdi3.o to the nm remove stage, we'll see how it goes on the next toolchain rebuild.

The callback function is defined to take a (void*) as parameter, but our templates let users use anything that fits inside sizeof(void*) to be passed in. Add pragmas to stop GCC warnings about this, since we already check the size of the type will fit in the allocated space.

Manually delete the divdi3.so from the libgcc.a library by running the updated EQT's 9.1-post script.

Exceptions are broken on all builds (GCC4.8-9.1) due to the removal of the PROGMEM non-32b read exception handler (added in the unstable pre3.0.0). Build the exception code with -mforce-l32 and patch accordingly to avoid LoadStore errors. Apply patches to select portions of the regex lib which use _stype_ (which is now in flash).

earlephilhower · 2020-06-09T19:08:28Z

@mcspr and @Jason2866 , I've hacked in the toolchain update in the PIO CI tests. Can you take a gander and see if there's something else you would suggest doing vs just overwriting the toolchain as the tests/platformio.sh does now.

mcspr · 2020-06-10T13:18:19Z

tests/platformio.sh

+    mv ~/.platformio/packages/toolchain-xtensa/package.json .save
+    rm -rf ~/.platformio/packages/toolchain-xtensa
+    mv $TRAVIS_BUILD_DIR/tools/xtensa-lx106-elf ~/.platformio/packages/toolchain-xtensa
+    mv .save ~/.platformio/packages/toolchain-xtensa/package.json
    python -c "import json; import os; fp=open(os.path.expanduser('~/.platformio/platforms/espressif8266/platform.json'), 'r+'); data=json.load(fp); data['packages']['framework-arduinoespressif8266']['version'] = '*'; fp.seek(0); fp.truncate(); json.dump(data, fp); fp.close()"


Regarding platform installation:

pio platform install has --without-package switch, this can avoid fetching the platform (toolchain-xtensa)

the same "*" glob for toolchain-extensa package described in the platform.json needs to be here?

Script itself uses pio ci down below, can't reach with github comment UI:

if get.py can instead of immediately fetching stuff just print the url, then "--board nodemcu_v2 --verbose" can be amended with the -O "platform_packages = toolchain-xtensa @ url" (shell-script quoting might become weird though). but, that would work only when package.json is inside the tarball / zip

manual fetch is also an option, package.json can just be generated (see root repo's package.json as an example)

Jason2866 · 2020-06-10T17:34:28Z

@earlephilhower time for some work load on your servers. 😆 Since you have already made the changes to include the package.json in your tarball generating script. So you could do it the way @mcspr has described. And i would like it a lot too because this gives everyone the possibilty to use the GCC 10.1 build chain easily with PlatformIO 👍☺
At the moment i manually add the package.json and upload on my github. Ugly...

earlephilhower · 2020-06-10T18:06:30Z

I'll adjust things to use the URL package once I get a chance to rebuild the toolchain. There were a couple of PIO issues squeaked out with the hacked version, though, which have been fixed.

Add a menu to enable GCC's built-in stack smash protection. When a subroutine goes past its end of stack, generate a crashdump on function exit like: ```` GCC detected stack overrun Stack corrupted, stack smash detected. >>>stack>>> ctx: cont sp: 3fffff20 end: 3fffffc0 offset: 0000 3fffff20: 40202955 00000001 0000001c 4020287e 3fffff30: feefeffe 000000fd 00000000 00000000 ... <<<stack<<< ```` Disabled by default because there is a small per-function code overhead (and CPU time if the function is called very frequently and is very small). BearSSL and LWIP are not built using stack smash detection, yet.

Report a fake exception to have the exception decoder print the actual faulting function. This won't tell you where in the function the issue happened, but it will tell you the function name first and foremost.

Gizmodo · 2020-06-20T21:50:54Z

@d-a-v When do you plan to merge PR? I can't wait for C++17

devyte · 2020-06-21T00:20:43Z

You can check the milestone for issues and PRs on the right. This is targeted for 3.0.0. That means it will be merged after 2.7.2, which is very close at hand.

d-a-v · 2020-06-21T06:23:23Z

@Gizmodo it is included in the alpha release

Jason2866 · 2020-06-21T07:38:07Z

@Gizmodo
I did a PlatformIO setup for testing the Project Tasmota with GCC10.1
You can use it with this entry in Platformio.ini
platform = https://github.com/Jason2866/platform-espressif8266.git#new_gcc
You can take a look at: https://github.com/Jason2866/Tasmota/tree/Gcc10.1

hreintke · 2020-07-03T14:35:50Z

@earlephilhower

There is a difference between the current an gcc9 toolchain in the .d files that are created.
These .d files are used by make in the build process.

Current :

T1.cpp.o: ..\T1.cpp ..\T1.h C:\temp\ArduinoC11\cores\esp8266/Arduino.h \
 C:/temp/ArduinoC11/tools/sdk/libc/xtensa-lx106-elf/include/stdlib.h \
 C:/temp/ArduinoC11/tools/sdk/libc/xtensa-lx106-elf/include/machine/ieeefp.h \
 C:/temp/ArduinoC11/tools/sdk/libc/xtensa-lx106-elf/include/_ansi.h \
 C:/temp/ArduinoC11/tools/sdk/libc/xtensa-lx106-elf/include/newlib.h \
 C:/temp/ArduinoC11/tools/sdk/libc/xtensa-lx106-elf/include/sys/config.h \
 C:/temp/ArduinoC11/tools/sdk/libc/xtensa-lx106-elf/include/sys/features.h \
 C:/temp/ArduinoC11/tools/sdk/libc/xtensa-lx106-elf/include/xtensa/config/core-isa.h \

gcc9

T1.cpp.o: ..\T1.cpp ..\T1.h C\:\temp\ArduinoC17\cores\esp8266/Arduino.h \
 C\:/temp/ArduinoC17/tools/sdk/libc/xtensa-lx106-elf/include/stdlib.h \
 C\:/temp/ArduinoC17/tools/sdk/libc/xtensa-lx106-elf/include/machine/ieeefp.h \
 C\:/temp/ArduinoC17/tools/sdk/libc/xtensa-lx106-elf/include/_ansi.h \
 C\:/temp/ArduinoC17/tools/sdk/libc/xtensa-lx106-elf/include/newlib.h \
 C\:/temp/ArduinoC17/tools/sdk/libc/xtensa-lx106-elf/include/sys/config.h \
 C\:/temp/ArduinoC17/tools/sdk/libc/xtensa-lx106-elf/include/sys/features.h \
 C\:/temp/ArduinoC17/tools/sdk/libc/xtensa-lx106-elf/include/xtensa/config/core-isa.h \

There is an additional \ after the first C

In my eclipse sloeber development environment that results in not being able to do incremental builds.

           Finished prerequisites of target file '..\T1.h'.
          No recipe for '..\T1.h' and no prerequisites actually changed.
          No need to remake target '..\T1.h'.
          Considering target file 'C\:\temp\ArduinoC17\cores\esp8266/Arduino.h'.
           File 'C\:\temp\ArduinoC17\cores\esp8266/Arduino.h' does not exist.
           Finished prerequisites of target file 'C\:\temp\ArduinoC17\cores\esp8266/Arduino.h'.
          Must remake target 'C\:\temp\ArduinoC17\cores\esp8266/Arduino.h'.
make: *** No rule to make target 'C\:\temp\ArduinoC17\cores\esp8266/Arduino.h', needed by 'T1.cpp.o'.  Stop.
"C:/temp/Sloeber/arduinoPlugin/tools/make/make --debug=verbose all" terminated with exit code 2. Build might be incomplete.

Is there a way to revert to the old format ?

earlephilhower · 2020-07-03T16:45:57Z

That's pretty strange. It's not a dependency format so much as an escaping of paths issue. I doubt there's anything you can do to make GCC not escape the colon in makefiles.

What make version do you have installed? Looking thru gnu.org I can see in gcc 4.x has an unescape_char(...,':') in the depfiles parser, but that is not present in the 3.x branch. No guarantees this fixes your problem, but it does look very promising.

hreintke · 2020-07-04T14:18:11Z

Sloeber eclipse has make 3.8.2 in it's distribution.
In contact with him now probably moving to 4.2.1 which doesn't have problems with the new format.

Thanks for pointing in the right direction.

earlephilhower · 2020-07-07T18:12:22Z

Here goes nothing...

mhightower83 · 2020-07-13T22:53:02Z

FWIW: The optimization does appear to have been improved. Code that constructs a slightly obscured divide by 0, as some of the examples I do to show the exception processing, may now report an HWDT event instead. In my case, the 10.1 compiler figured it out and replaced it with a break 1, 15. My workaround was to add a noinline to my local divide a / b function that I call to generate the exception.

Upgrade to GCC 9.1 toolchain

59023e9

earlephilhower added 10 commits July 13, 2019 08:18

Rebuilt using pure GNU binutils and GCC

8642204

Remove dependencies on earlier forked GNU utilities (gcc-xtensa, binutils-gdb-xtensa) and just use GCC sources, unmodified (except for patches in the esp-quick-toolchain directories).

Rebuild bearssl using new toolchain

354cfcd

Fix GDBstub linkage options

99a0c6f

GDB works with pure GNU GCC and pure GNU binutils now. Still warnings galore, but tested with the example sketch in the docs.

Fix digitalRead alias warning

83f308e

Remove gdb stub warnings w/a pragma

33fee67

Merge branch 'master' into gcc9.1

726f0e9

Fix AxTLS alias function defs to match real code

ef74b7c

Fix WiFiClientSecure implicit default copy ctor

a15ab56

These both use shared-ptrs to handle refcnts to allocated data, so using the default copy constructor is fine (and has been in use for a long time).

Dummy size for heap to avoid GCC 8/9 warnings

3065a3f

Make GCC think _heap_start is large enough to avoid the basic (and incorrect) bounds-checking warnings it produces. The size chosen is arbitrary and does not affect the actual size of the heap in any way.

earlephilhower requested a review from d-a-v July 13, 2019 17:43

Make heap an undefined extend array

a68d8ce

Instead of a bogus size, use an indefinite size for the heap to avoid GCC warnings

earlephilhower added 2 commits July 13, 2019 19:56

Trivial tab to space fix

b0a5fe6

Update SDFat to remove FatFile warnings

bb3df45

earlephilhower force-pushed the gcc9.1 branch from a78174b to bb3df45 Compare July 14, 2019 03:30

earlephilhower added 3 commits July 13, 2019 21:06

Remove GCC support fcn that's in ROM

f4cc2d1

Manually delete the divdi3.so from the libgcc.a library by running the updated EQT's 9.1-post script.

Merge branch 'master' into gcc9.1

1168360

d-a-v approved these changes Jul 15, 2019

View reviewed changes

earlephilhower force-pushed the gcc9.1 branch 2 times, most recently from 40e26f1 to d04816e Compare July 15, 2019 18:19

earlephilhower force-pushed the gcc9.1 branch from d04816e to 1cb9282 Compare July 15, 2019 19:26

earlephilhower added 2 commits July 15, 2019 12:30

Rebuild Bearssl using latest GCC push

e5ee285

Merge branch 'master' into gcc9.1

86812e6

mcspr reviewed Jun 10, 2020

View reviewed changes

Fix several asm warnings in PIO build

c4dea28

earlephilhower force-pushed the gcc9.1 branch from 32b1d56 to c4dea28 Compare June 10, 2020 17:54

earlephilhower added 4 commits June 16, 2020 18:39

Merge branch 'master' into gcc9.1

8f8a196

Fix duplicated stc=gnu99/c17 in build

2c99453

Dump faulting function PC in stack overflow

5559729

Report a fake exception to have the exception decoder print the actual faulting function. This won't tell you where in the function the issue happened, but it will tell you the function name first and foremost.

Rebuild with Platform.io JSON tag in release tgzs

db8ef89

earlephilhower force-pushed the gcc9.1 branch from 4b236ea to db8ef89 Compare July 7, 2020 03:52

d-a-v added 2 commits July 7, 2020 12:08

Merge branch 'master' into gcc9.1

d17dc0f

Merge branch 'master' into gcc9.1

837a4e0

earlephilhower merged commit d979b57 into esp8266:master Jul 7, 2020

mcspr mentioned this pull request Jul 23, 2020

Build errors xtensa-lx106-elf-gcc: error: unrecognized command line option '-std=c17' platformio/platform-espressif8266#224

Closed

mcspr mentioned this pull request Jul 31, 2020

Cannot use generic platform_packages = $URL with different OS platformio/platformio-core#3612

Closed

mcspr mentioned this pull request Sep 27, 2020

test: use -std=gnu17 instead of -std=c17 #7588

Merged

earlephilhower deleted the gcc9.1 branch September 27, 2020 05:54

hreintke mentioned this pull request May 25, 2021

Compilation speed of ESP8266 core 3.0.0 #8069

Closed

mcspr mentioned this pull request Jul 28, 2023

optionally move float emulation code into iram #8958

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Upgrade to GCC 10.1 toolchain #6294

Upgrade to GCC 10.1 toolchain #6294

earlephilhower commented Jul 13, 2019 •

edited

Loading

earlephilhower commented Jul 13, 2019

earlephilhower commented Jul 13, 2019

d-a-v commented Jul 13, 2019 •

edited

Loading

earlephilhower commented Jul 14, 2019 •

edited

Loading

earlephilhower commented Jun 9, 2020

mcspr Jun 10, 2020

Jason2866 commented Jun 10, 2020 •

edited

Loading

earlephilhower commented Jun 10, 2020

Gizmodo commented Jun 20, 2020

devyte commented Jun 21, 2020

d-a-v commented Jun 21, 2020

Jason2866 commented Jun 21, 2020 •

edited

Loading

hreintke commented Jul 3, 2020

earlephilhower commented Jul 3, 2020

hreintke commented Jul 4, 2020

earlephilhower commented Jul 7, 2020

mhightower83 commented Jul 13, 2020 •

edited

Loading

Upgrade to GCC 10.1 toolchain #6294

Upgrade to GCC 10.1 toolchain #6294

Conversation

earlephilhower commented Jul 13, 2019 • edited Loading

earlephilhower commented Jul 13, 2019

earlephilhower commented Jul 13, 2019

d-a-v commented Jul 13, 2019 • edited Loading

earlephilhower commented Jul 14, 2019 • edited Loading

earlephilhower commented Jun 9, 2020

mcspr Jun 10, 2020

Choose a reason for hiding this comment

Jason2866 commented Jun 10, 2020 • edited Loading

earlephilhower commented Jun 10, 2020

Gizmodo commented Jun 20, 2020

devyte commented Jun 21, 2020

d-a-v commented Jun 21, 2020

Jason2866 commented Jun 21, 2020 • edited Loading

hreintke commented Jul 3, 2020

earlephilhower commented Jul 3, 2020

hreintke commented Jul 4, 2020

earlephilhower commented Jul 7, 2020

mhightower83 commented Jul 13, 2020 • edited Loading

earlephilhower commented Jul 13, 2019 •

edited

Loading

d-a-v commented Jul 13, 2019 •

edited

Loading

earlephilhower commented Jul 14, 2019 •

edited

Loading

Jason2866 commented Jun 10, 2020 •

edited

Loading

Jason2866 commented Jun 21, 2020 •

edited

Loading

mhightower83 commented Jul 13, 2020 •

edited

Loading