# The firmware you flash

The same call is a different function on every board. Context is the image this product runs.

[Book a demo](https://byteask.cal.com/pratyush/book-a-demo) [See the evidence](https://byteask.ai/industries/embedded#evidence)

## The build gives the context

ByteAsk Context indexes the functions and calls your compile commands select. The examples below show what was captured for this domain.

portENTER_CRITICAL();

Text search

205 lines in 168 files. 128 #defines, in 128 port files. GCC 42, IAR 32, RVDS/Keil 7, Renesas 6, and others

POSIX index

One definition: vPortEnterCritical(), portable/ThirdParty/GCC/Posix/port.c:249, with 32 callers

Lines and #defines: rg -w portENTER_CRITICAL, FreeRTOS-Kernel 8e2dd5b. The definition and the 32 callers are the POSIX build index, from the embedded deck notes, not from that rg.

[How ByteAsk Context works](https://byteask.ai/context)

## A watchpoint finds the wrong joint

armctl, a simulated arm controller. The watchpoint stops on the joint written by mistake. The rebuilt program finishes with every joint inside its soft limits.

Shown here: a watchpoint in gdb, then a rebuild. This recording does not run the sanitizers after the fix.

```
Hardware watchpoint 2: -location *(double*)0x555555559280
Old value = -1.9057374359644947
New value = -1.8236172821972838
$8 = 5
$12 = "elbow\000\000"
I have the concrete bad store: after reprioritization, controller `joint_id_=5` still targets `slot_=5`, and slot 5 now holds `elbow`.
+            for (int i = 0; i < kNumJoints; ++i) ctl[i].bind(static_cast<std::uint32_t>(i));
completed, all joints within soft limits
```

Recorded 2026-09-11.

Read the full captured excerpt

```
{"action":"cmd","session":1,"command":"print &g_joints[5].measured_rad"}
$6 = (double *) 0x555555559280 <rtc::g_joints+256>
{"action":"cmd","session":1,"command":"watch -location *(double*)0x555555559280"}
Hardware watchpoint 2: -location *(double*)0x555555559280
{"action":"cmd","session":1,"command":"continue"}
Old value = -1.9057374359644947
New value = -1.8236172821972838
{"action":"cmd","session":1,"command":"print this->joint_id_"}
$8 = 5
{"action":"cmd","session":1,"command":"print g_joints[this->slot_].name"}
$12 = "elbow\000\000"
I have the concrete bad store: after reprioritization, controller `joint_id_=5` still targets `slot_=5`, and slot 5 now holds `elbow`.
+            for (int i = 0; i < kNumJoints; ++i) ctl[i].bind(static_cast<std::uint32_t>(i));
/usr/bin/zsh -lc './build/armctl 20000' in ~/byteask/cppcon/demos/04-runtime-forensics
completed, all joints within soft limits
```

## The source behind the answer

STM32F405/415, STM32F407/417, STM32F427/437 and STM32F429/439 Advanced Arm-based 32-bit MCUs Reference Manual (RM0090, DM00031020)
pp.613–614

> Figure 169 gives an example of counter behavior when TI1FP1 polarity is inverted (same configuration as above except CC1P=1).

Retrieved with `search_docs` on 2026-10-26. This document lookup is separate from the recorded session above.

## On your build

[Book a demo](https://byteask.cal.com/pratyush/book-a-demo)

[On-prem and air-gapped](https://byteask.ai/enterprise)

---
HTML version: https://byteask.ai/industries/embedded