-
Notifications
You must be signed in to change notification settings - Fork 109
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Desktop support (Linux / Windows / Mac) (ESF-122) #102
base: master
Are you sure you want to change the base?
Conversation
Update: firmware flashing works |
examples/common/example_common.c
Outdated
@@ -336,7 +346,8 @@ esp_loader_error_t load_ram_binary(const uint8_t *bin) | |||
printf("Start loading\n"); | |||
esp_loader_error_t err; | |||
const esp_loader_bin_header_t *header = (const esp_loader_bin_header_t *)bin; | |||
esp_loader_bin_segment_t segments[header->segments]; | |||
//esp_loader_bin_segment_t segments[header->segments]; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why is this being moved to the heap?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Because it didn't compile on MSVC
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah yes, after C99 VLAs are a GNU extension as well. We shouldn't be using VLAs in the first place IMO but I am unsure about using malloc()
, as we might want to support platforms in the future that don't have it available. Also, using malloc()
in example code for embedded platforms where malloc()
is often prohibited is not ideal, especially as some of these example functions happen to be used in production a lot of the time.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A simple static array with a max segment count should suffice.
@@ -295,7 +295,7 @@ esp_loader_error_t loader_spi_parameters(uint32_t total_size) | |||
return send_cmd(&spi_cmd, sizeof(spi_cmd), NULL); | |||
} | |||
|
|||
__attribute__ ((weak)) void loader_port_debug_print(const char *str) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Probably leftover code from testing, a weak symbol shouldn't hurt anything.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@DNedic I think the reason is that __attribute__((weak))
syntax is a GNU extension, and is not supported by MSVC compiler. Same for packed
— MSCV has a different way of specifying the same.
I think this will still need to be cleaned up to move all the compiler-specific things into a common header.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I agree, we should move these platform details into a separate header if we merge this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since LLVM GNU extensions, is there a way to build for Windows using LLVM and just not bother with MSVC? As far as I've seen the incompatibilities go deeper than GNU extensions.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Mingw might work, but I'd prefer an "if msvc else" precompiler flag. Is does actually compile but the problem is this symbol name is being used multiple times and thus the linker complains. Renaming one of them would also fix the issue.
port/serial_port.h
Outdated
} loader_serial_config_t; | ||
|
||
esp_loader_error_t loader_port_serial_init(const loader_serial_config_t *config); | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Perhaps a deinit function would be useful?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Agreed
.gitmodules
Outdated
@@ -1,3 +1,6 @@ | |||
[submodule "submodules/stm32-cmake"] | |||
path = submodules/stm32-cmake | |||
url = https://github.com/ObKo/stm32-cmake.git | |||
[submodule "submodules/serial"] | |||
path = submodules/serial | |||
url = https://github.com/wjwwood/serial |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just my 2c, I would recommend fetching this project at build time using ExternalProject
or FetchContent
CMake modules, instead of adding a submodule into esp-serial-flasher.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Good idea! This would be a good time to move over the stm32-cmake
submodule to FetchContent
as well.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Perhaps a find_package script which can do the fetching if needed. In my application the serial library is already present. As a note, I didn't use add_subdirectory on this since it wants catkin as a dependancy, which I don't need here.
Hello @electromuis , nice work and thank you for the contribution! I've left some initial comments, most of them are probably due to this being WIP. |
Thanks for the comments, added my replies. The README could also be updated since "from other host microcontrollers" doesn't count for a big desktop CPU haha |
Tried to address each feedback point |
@@ -1,8 +1,10 @@ | |||
# esp-serial-flasher | |||
|
|||
`esp-serial-flasher` is a portable C library for flashing or loading apps to RAM of Espressif SoCs from other host microcontrollers. | |||
`esp-serial-flasher` is a portable C library for flashing or loading apps to RAM of Espressif SoCs from other host that are supported through their dedicated port drivers. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Perhaps it's better to be upfront here. I would preffer something like:
`esp-serial-flasher` is a portable C library for flashing or loading apps to RAM of Espressif SoCs from other microcontrollers, SBCs and even desktop PCs.
return; | ||
} | ||
|
||
// This might be the best we can do without GPIO |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please take a look at how esptool
does this. There are several reset strategies depending on whether the host is Windows/Linux or MacOS.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The wjwwood serial library seems to have support for setting DTR/RTS
https://github.com/wjwwood/serial/blob/69e0372cf0d3796e84ce9a09aff1d74496f68720/src/serial.cc#L404-L406
Questions is whether setting them independently is fast enough. From my experience: most likely not.
I've used the serial class from Qt which has similar abstractions and still had to resort to OS specific calls to switch DTR/RTS at the same time.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Isn't that what @electromuis is doing here? The question I was addressing is whether the platform specifics have propagated to the library itself. If not, this code is good as is.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think what he means is dtr and rts would need to switch in a single call instead of the two -> set calls listed currently.
The board I'm using currently doesn't even reset with esptool.py so hard to test
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think what he means is dtr and rts would need to switch in a single call instead of the two -> set calls listed currently.
The board I'm using currently doesn't even reset with esptool.py so hard to test
Yes that's what I've ment.
For a board I made using the esp32s3 I also have a desktop client written in c++.
It would be great if users could update the firmware directly from the tool. Wanting to keep the tool small by not including a python install to use esptool, I found this library to contain the most important parts.
I've done some work to make it work on x86 systems by using the https://github.com/wjwwood/serial library.
It's not completed yet and could probably use a good eye but I wanted to share it at least so other interested people could find it.
A new example is added called: desktop_esp32_example. It builds and rust on my windows system. Have not tried flashing to an actual board yet, will have to check got esptool does the resetting since normally a PC wouldn't have gpio access.