exception.what() not a string

[mhightower83]

Basic Infos

• This issue complies with the issue POLICY doc.
• I have read the documentation at readthedocs and the issue is not addressed there.
• I have tested that the issue is present in current master branch (aka latest git).
• I have searched the issue tracker for a similar issue.
• If there is a stack dump, I have decoded it.
• I have filled out all fields below.

Platform

• Hardware: [ESP-01S]
• Core Version: [80bf716]
• Development Env: [Arduino IDE]
• Operating System: [Ubuntu]

Settings in IDE

• Module: [Generic ESP8266 Module]
• Flash Mode: [dout]
• Flash Size: [1MB]
• lwip Variant: [v2 Lower Memory]
• Reset Method: [nodemcu]
• Flash Frequency: [40Mhz]
• CPU Frequency: [80Mhz]
• Upload Using: [SERIAL]
• Upload Speed: [921600 (serial upload only)

Problem Description

For the case of a missing try/catch around new, a secondary crash within postmortem occurs while processing the unhandled C++ Exception.

The second crash is caused when e.what() returns a value of 0x12.

Arduino/cores/esp8266/core_esp8266_main.cpp

Line 296 in 80bf716

__unhandled_exception( e.what() );

resulting in a new ESP exception in postmortem at

Arduino/cores/esp8266/core_esp8266_postmortem.cpp

Line 144 in 80bf716

ets_printf_P(PSTR("\nUnhandled C++ exception: %S\n"), s_unhandled_exception);

MCVE Sketch

/*
   Build with Tools->C++ Exceptions: "enable"
*/
char *p = NULL;

void setup() {
  Serial.begin(115200);
  delay(100);
  Serial.printf("\nUp and running ...\n");

  // loops - crashing in postmortem
  p = new char[128000];

  if (p) {
    Serial.printf("\n%p, \"new\" worked, it should not have!\n", p);
  }
}

void loop() {}

Debug Messages

Up and running ...

User exception (panic/abort/assert)
--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

--------------- CUT HERE FOR EXCEPTION DECODER ---------------

 ets Jan  8 2013,rst cause:4, boot mode:(3,7)

wdt reset

[mcspr]

... #8717 as an alternative way of overloading new. note c++17 also has aligned ones, seems like a waste writing all of those yet again :/

std::bad_alloc::what() is indeed returning 0x12.

arduino8716 [1]> xtensa-lx106-elf-gdb --batch --ex "disassemble 'std::bad_alloc::what'" --ex "x 0x40204674" .pio/build/d1_mini/firmware.elf
Dump of assembler code for function std::bad_alloc::what() const:
   0x4020d614 <+0>:     l32r    a2, 0x40204674
   0x4020d617 <+3>:     ret.n
End of assembler dump.
0x40204674:     0x00000012

Here's the string, though

> xtensa-lx106-elf-objdump -s -j .irom.exceptiontext  ~/.platformio/packages/framework-arduinoespressif8266@src-31d658a59f41540201fc3726a1394910/tools/sdk/lib/libstdc++-exc.a
...
bad_alloc.o:     file format elf32-xtensa-le

Contents of section .irom.exceptiontext:
 0000 7374643a 3a626164 5f616c6c 6f6300    std::bad_alloc.
...

(see toolchain .patch)

[mcspr]

this seems to fix the discarded address, just not sure why the rule(s) below try to put things in .text

diff --git a/tools/sdk/ld/eagle.app.v6.common.ld.h b/tools/sdk/ld/eagle.app.v6.common.ld.h
index 77c834ae..ed838918 100644
--- a/tools/sdk/ld/eagle.app.v6.common.ld.h
+++ b/tools/sdk/ld/eagle.app.v6.common.ld.h
@@ -181,8 +181,8 @@ SECTIONS
     *libgcc.a:_fixsfsi.o(.literal .text)
     *libgcc.a:_cmpdf2.o(.literal .text)
     *libgcc.a:_cmpsf2.o(.literal .text)
     *libstdc++.a:( .literal .text .literal.* .text.*)
-    *libstdc++-exc.a:( .literal .text .literal.* .text.*)
+    *libstdc++-exc.a:( .literal .text .literal.* .text.* .irom.exceptiontext)
     *libsmartconfig.a:(.literal .text .literal.* .text.*)
     *liblwip_gcc.a:(.literal .text .literal.* .text.*)
     *liblwip_src.a:(.literal .text .literal.* .text.*)

[mhightower83]

@mcspr
Interesting options.
Your change to eagle.app.v6.common.ld.h worked for me.
Putting the strings into .irom0.text appears to land in the PROGMEM address range.
Thanks!

[mhightower83]

@mcspr Also for me, removing the comma here allows things to work.

Arduino/tools/sdk/ld/eagle.app.v6.common.ld.h

Line 223 in 80bf716

*(.irom.exceptiontext, .rodata.exceptiontext)

[mcspr]

(somewhat late comment, thanks for the fix!)

^ re above and Earles comment about ld, there is actually nothing about such syntax option in up-to-date binutils documentation.

https://ftp.gnu.org/old-gnu/Manuals/ld-2.9.1/html_chapter/ld_toc.html

January 1994

vs.
https://sourceware.org/binutils/docs/ld/Input-Section-Basics.html

Nothing in the testsuite either; git log -p -- ./ld/testsuite/**.ld, search for , . gives nothing useful. Noticed this 'ordered' list, but not really sure of the difference between the two (manual sorting order?)

+  .got            : { *(.got.plt) *(.igot.plt) *(.got) *(.igot) }

https://sourceware.org/git?p=binutils-gdb.git;a=tree;f=ld/testsuite;h=1f226f4c011ade55ee276991a21631a19ef675fd;hb=HEAD