Skip to main content

Symbol Files & Build Ids


In order to fully process uploaded data, such as coredumps, Memfault needs to locate the Symbol File (ELF) that corresponds to the software that produced the uploaded data. Without an exact match, Memfault will not be able to decode the uploaded data. Processing will still happen, but the resulting Trace will contain limited information until the symbol file is uploaded and the data is reprocessed.

To that end, the SDK adds metadata to the uploaded data to identify the software that produced the data. Likewise, any Symbol File that is uploaded to Memfault must have that same identifying metadata.

Identifier Types

There are 3 types of identifiers:

Memfault Build IdGNU Build IdSoftware Type & Version
Automatically generated hash
Manually chosen
Compiler requirementGNU/gcc
ID embedded inside ELF
ID passed manually upon Symbol File upload
Requires running Python script to post-process ELF

We recommend using either the Memfault Build Id or GNU Build Id option. The Build Ids are automatically generated based on the contents of the ELF file.

Software Type & Version identifiers are manually assigned and therefore error-prone.


When uploading a Symbol File (ELF) that has a Build Id embedded, it is possible to add a Software Type & Version based identifier, in addition to the embedded Build Id. This will link the Symbol File with the Software Version. That way, you can later look up and download the Symbol File by Software Type and Version through the Software Versions UI in Memfault.

Note that for the purpose of finding a Symbol File to match uploaded data, the Build Id always takes precedence over Software Type & Version.

Adding Build Ids

Memfault has two ways to add a Build ID to a target. Select the appropriate one below based on your toolchain setup.


Requires use of the GNU GCC compiler.

  1. Add -Wl,--build-id to flags passed to linker via GCC
  2. Add #define MEMFAULT_USE_GNU_BUILD_ID 1 to third_party/memfault/memfault_platform_config.h
  3. Add the following snippet to your projects linker script (.ld file) where "<YOUR_FLASH_SECTION>" below will match the name of the MEMORY section that read-only data and text is placed in. :
__start_gnu_build_id_start = .;

Be sure to update <YOUR_FLASH_SECTION> to match the name of the section .text is placed in!

The output section name is used to locate the build id when the symbol file is uploaded. The __start_gnu_build_id_start identifier is used at compile time by the SDK for populating the build id.

See the FreeRTOS example app in the Memfault SDK for a reference implementation!

Uploading Symbol Files using Memfault CLI

Please follow the instructions to install the Memfault CLI tool first.

Uploading Symbol File with Build Id

Because the Build Id is embedded inside the ELF file, the Build Id is automatically recognized and extracted and does not need to be passed as an argument to the memfault tool:

memfault --org-token $ORG_TOKEN --org acme-inc --project smart-sink \
upload-mcu-symbols \
If you are using Link-Time Optimization in your build, you may get an error when uploading the symbol file

Using link-time optimization (-flto) can cause the symbol upload to fail:

ERROR: main-with-lto.elf: Build Id missing. Specify --software-version and
--software-type options or add a Build Id (see
Usage: memfault upload-mcu-symbols [OPTIONS] PATH

Error: Upload failed!

-flto can inline the contents of the Build ID variable, removing the symbol from the executable. To work around this, add this flag to the linker args:


Uploading Symbol File with Software Type & Version

To associate a Software Type & Software Version with a Symbol File, the --software-type and --software-version arguments need to be passed to the memfault tool.

Optionally, a --revision (Git commit hash, SVN revision, etc) can be passed as well to store the version control revision in the newly created Software Version.

memfault --org-token $ORG_TOKEN --org acme-inc --project smart-sink \
upload-mcu-symbols \
--software-type stm32-fw \
--software-version 1.0.0-alpha \
--revision 89335ffade90ff7697e2ce5238bd4c68978b6d6e \

Frequently Asked Questions

When using a Build ID, is it okay if there is not always a 1:1 mapping with Software Version?

Yes, in fact this is one of the benefits of using Build IDs. In this scenario there will always be a 1:1 mapping between the binary running and a Build ID so Memfault is guaranteed to recover accurate stack traces while at the same time giving you flexibility around how you report software versions.

Two common situations where there is not a 1:1 mapping are the following:

  • "Dev" builds -- Memfault recommends using semver and for pre-release debug builds, using a common software version (i.e 1.1.1-dev). This way all the crashes and issues observed during debug get bucketed under a single version that's easy to track in the UI. In this situation you wind up with many separate Build ID mapping to a single Software Version
  • In some systems, multiple Software Versions can map to the same Symbol File (via Build ID), because the Software Version in the device is not stored within the application binary, so it does not impact the Symbol File (and corresponding Build ID).