|
1 |
| -# Build failures |
| 1 | +# Troubleshooting build failures |
2 | 2 |
|
3 |
| -This page lists common reasons why resolution and installation fails with a build error and how to |
4 |
| -fix them. |
| 3 | +uv needs to build packages when there is not a compatible wheel (a pre-built distribution of the |
| 4 | +package) available. Building packages can fail for many reasons. |
5 | 5 |
|
6 |
| -### Why does uv build a package? |
| 6 | +## Recognizing a build failure |
| 7 | + |
| 8 | +An example build failure can be produced by trying to install and old version of numpy on a new, |
| 9 | +unsupported version of Python: |
| 10 | + |
| 11 | +```console |
| 12 | +$ uv pip install -p 3.13 'numpy<1.20' |
| 13 | +Resolved 1 package in 62ms |
| 14 | + × Failed to download and build `numpy==1.19.5` |
| 15 | + ├─▶ Build backend failed to determine requirements with `build_wheel()` (exit status: 1) |
| 16 | + |
| 17 | + │ [stderr] |
| 18 | + │ Traceback (most recent call last): |
| 19 | + │ File "<string>", line 8, in <module> |
| 20 | + │ from setuptools.build_meta import __legacy__ as backend |
| 21 | + │ File "/Users/example/.cache/uv/builds-v0/.tmp96A0WB/lib/python3.13/site-packages/setuptools/__init__.py", line 9, in <module> |
| 22 | + │ import distutils.core |
| 23 | + │ ModuleNotFoundError: No module named 'distutils' |
| 24 | + |
| 25 | + ╰─▶ distutils was removed from the standard library in Python 3.12. Consider adding a constraint |
| 26 | + (like `numpy >1.19.5`) to avoid building a version of numpy that depends on distutils. |
| 27 | +``` |
| 28 | + |
| 29 | +Notice that the error message is prefaced by "Build backend failed to determine requirements". |
| 30 | + |
| 31 | +The build failure includes the `[stderr]` (and `[stdout]`, if present) from the build backend that |
| 32 | +was used for the build. The error logs are not from uv itself. |
| 33 | + |
| 34 | +The message following the `╰─▶` is a hint provided by uv, to help resolve common build failures. A |
| 35 | +hint will not be available for all build failures. |
| 36 | + |
| 37 | +## Confirming that a build failure is specific to uv |
| 38 | + |
| 39 | +Build failures are usually related to your system and the build backend. It is rare that a build |
| 40 | +failure is specific to uv. You can confirm that the build failure is not related to uv by attempting |
| 41 | +to reproduce it with pip: |
| 42 | + |
| 43 | +```console |
| 44 | +$ uv venv -p 3.13 --seed |
| 45 | +$ source .venv/bin/activate |
| 46 | +$ pip install --use-pep517 'numpy==1.19.5' |
| 47 | +Collecting numpy==1.19.5 |
| 48 | + Using cached numpy-1.19.5.zip (7.3 MB) |
| 49 | + Installing build dependencies ... done |
| 50 | + Getting requirements to build wheel ... done |
| 51 | +ERROR: Exception: |
| 52 | +Traceback (most recent call last): |
| 53 | + ... |
| 54 | + File "/Users/example/.cache/uv/archive-v0/3783IbOdglemN3ieOULx2/lib/python3.13/site-packages/pip/_vendor/pyproject_hooks/_impl.py", line 321, in _call_hook |
| 55 | + raise BackendUnavailable(data.get('traceback', '')) |
| 56 | +pip._vendor.pyproject_hooks._impl.BackendUnavailable: Traceback (most recent call last): |
| 57 | + File "/Users/example/.cache/uv/archive-v0/3783IbOdglemN3ieOULx2/lib/python3.13/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py", line 77, in _build_backend |
| 58 | + obj = import_module(mod_path) |
| 59 | + File "/Users/example/.local/share/uv/python/cpython-3.13.0-macos-aarch64-none/lib/python3.13/importlib/__init__.py", line 88, in import_module |
| 60 | + return _bootstrap._gcd_import(name[level:], package, level) |
| 61 | + ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
| 62 | + File "<frozen importlib._bootstrap>", line 1387, in _gcd_import |
| 63 | + File "<frozen importlib._bootstrap>", line 1360, in _find_and_load |
| 64 | + File "<frozen importlib._bootstrap>", line 1310, in _find_and_load_unlocked |
| 65 | + File "<frozen importlib._bootstrap>", line 488, in _call_with_frames_removed |
| 66 | + File "<frozen importlib._bootstrap>", line 1387, in _gcd_import |
| 67 | + File "<frozen importlib._bootstrap>", line 1360, in _find_and_load |
| 68 | + File "<frozen importlib._bootstrap>", line 1331, in _find_and_load_unlocked |
| 69 | + File "<frozen importlib._bootstrap>", line 935, in _load_unlocked |
| 70 | + File "<frozen importlib._bootstrap_external>", line 1022, in exec_module |
| 71 | + File "<frozen importlib._bootstrap>", line 488, in _call_with_frames_removed |
| 72 | + File "/private/var/folders/6p/k5sd5z7j31b31pq4lhn0l8d80000gn/T/pip-build-env-vdpjme7d/overlay/lib/python3.13/site-packages/setuptools/__init__.py", line 9, in <module> |
| 73 | + import distutils.core |
| 74 | +ModuleNotFoundError: No module named 'distutils' |
| 75 | +``` |
| 76 | + |
| 77 | +!!! important |
| 78 | + |
| 79 | + The `--use-pep517` flag should be included with the `pip install` invocation to ensure the same |
| 80 | + build isolation behavior. uv always uses [build isolation by default](https://docs.astral.sh/uv/pip/compatibility/#pep-517-build-isolation). |
| 81 | + |
| 82 | + We also recommend including the `--force-reinstall` and `--no-cache` options when reproducing |
| 83 | + failures. |
| 84 | + |
| 85 | +Since this build failure occurs in pip too, it is not likely to be a bug with uv. |
| 86 | + |
| 87 | +If a build failure is reproducible with another installer, you should file an issue upstream (in |
| 88 | +this example, with `numpy` or `setuptools`), find a way to avoid building the package in the first |
| 89 | +place, or make the necessary adjustments to your system for the build to succeed. |
| 90 | + |
| 91 | +## Why does uv build a package? |
7 | 92 |
|
8 | 93 | When generating the cross-platform lockfile, uv needs to determine the dependencies of all packages,
|
9 | 94 | even those only installed on other platforms. uv tries to avoid package builds during resolution. It
|
10 | 95 | uses any wheel if exist for that version, then tries to find static metadata in the source
|
11 | 96 | distribution (mainly pyproject.toml with static `project.version`, `project.dependencies` and
|
12 |
| -`project.optional-dependencies` or METADATA of at least version 2.2). Only if all of that fails, it |
13 |
| -builds the package. |
| 97 | +`project.optional-dependencies` or METADATA v2.2+). Only if all of that fails, it builds the |
| 98 | +package. |
14 | 99 |
|
15 | 100 | When installing, uv needs to have a wheel for the current platform for each package. If no matching
|
16 | 101 | wheel exists in the index, uv tries to build the source distribution.
|
17 | 102 |
|
18 | 103 | You can check which wheels exist for a PyPI project under “Download Files”, e.g.
|
19 | 104 | https://pypi.org/project/numpy/2.1.1/#files. Wheels with `...-py3-none-any.whl` filenames work
|
20 |
| -everywhere, others have the operating system and platform in the filename. For the linked numpy |
21 |
| -version, you can see that Python 3.10 to 3.13 on MacOS, Linux and Windows are supported. |
22 |
| - |
23 |
| -### Fixes and Workarounds |
24 |
| - |
25 |
| -- If the build error mentions a missing header or library, there is often a matching package in your |
26 |
| - system package manager. |
27 |
| - |
28 |
| - Example: When `uv pip install mysqlclient==2.2.4` fails on Ubuntu, you need to run |
29 |
| - `sudo apt install default-libmysqlclient-dev build-essential pkg-config` to install the MySQL |
30 |
| - headers ([https://pypi.org/project/mysqlclient/2.2.4/](https://pypi.org/project/mysqlclient/2.2.4/#Linux)) |
31 |
| - |
32 |
| -- If the build error mentions a failing import, consider |
33 |
| - [deactivating build isolation](https://docs.astral.sh/uv/concepts/projects/#build-isolation). |
34 |
| -- If a package fails to build during resolution and the version that failed to build is older than |
35 |
| - the version you want to use, try adding a |
36 |
| - [constraint](https://docs.astral.sh/uv/reference/settings/#constraint-dependencies) with a lower |
37 |
| - bound (e.g. `numpy>=1.17`). Sometimes, due to algorithmic limitations, the uv resolver tries to |
38 |
| - find a fitting version using unreasonably old packages, which can be prevented by using lower |
39 |
| - bounds. |
40 |
| -- Consider using a different Python version for locking and/or installation (`-p`). If you are using |
41 |
| - an older Python version, you may need to use an older version of certain packages with native code |
42 |
| - too, especially for scientific code. Example: torch 1.12.0 support Python 3.7 to 3.10 |
43 |
| - (https://pypi.org/project/torch/1.12.0/#files), while numpy 2.1.0 supports Python 3.10 to 3.13 |
44 |
| - (https://numpy.org/doc/stable/release/2.1.0-notes.html#numpy-2-1-0-release-notes), so both |
45 |
| - together mean you need Python 3.10 (or upgrade torch). |
46 |
| -- If locking fails due to building a package from a platform you do not support, consider |
47 |
| - [declaring resolver environments](https://docs.astral.sh/uv/reference/settings/#environments) with |
48 |
| - your supported platforms. |
49 |
| -- If you support a large range of Python versions, consider using markers to use older versions for |
50 |
| - older Python versions and newer versions for newer Python version. In the example, numpy tends to |
51 |
| - support four Python minor version at a time, so to support Python 3.8 to 3.13, the versions need |
52 |
| - to be split: |
53 |
| - |
54 |
| - ``` |
55 |
| - numpy>=1.23; python_version >= "3.10" |
56 |
| - numpy<1.23; python_version < "3.10" |
57 |
| - ``` |
58 |
| - |
59 |
| -- If locking fails due to building a package from a different platform, as an escape hatch you can |
60 |
| - [provide dependency metadata manually](https://docs.astral.sh/uv/reference/settings/#dependency-metadata). |
61 |
| - As uv can not verify this information, it is important to specify correct metadata in this |
62 |
| - override. |
| 105 | +everywhere, others have the operating system and platform in the filename. In the linked `numpy` |
| 106 | +example, you can see that there are pre-built distributions for Python 3.10 to 3.13 on MacOS, Linux |
| 107 | +and Windows. |
| 108 | + |
| 109 | +## Common build failures |
| 110 | + |
| 111 | +The following examples demonstrate common build failures and how to resolve them. |
| 112 | + |
| 113 | +### Command is not found |
| 114 | + |
| 115 | +If the build error mentions a missing command, for example, `gcc`: |
| 116 | + |
| 117 | +<!-- docker run --platform linux/x86_64 -it ghcr.io/astral-sh/uv:python3.10-bookworm-slim /bin/bash -c "uv pip install --system pysha3==1.0.2" --> |
| 118 | + |
| 119 | +```hl_lines="17" |
| 120 | +× Failed to download and build `pysha3==1.0.2` |
| 121 | +╰─▶ Build backend failed to build wheel through `build_wheel` (exit status: 1) |
| 122 | +
|
| 123 | + [stdout] |
| 124 | + running bdist_wheel |
| 125 | + running build |
| 126 | + running build_py |
| 127 | + creating build/lib.linux-x86_64-cpython-310 |
| 128 | + copying sha3.py -> build/lib.linux-x86_64-cpython-310 |
| 129 | + running build_ext |
| 130 | + building '_pysha3' extension |
| 131 | + creating build/temp.linux-x86_64-cpython-310/Modules/_sha3 |
| 132 | + gcc -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -fPIC -DPY_WITH_KECCAK=1 -I/root/.cache/uv/builds-v0/.tmpxAJdUa/include -I/usr/local/include/python3.10 -c Modules/_sha3/sha3module.c -o |
| 133 | + build/temp.linux-x86_64-cpython-310/Modules/_sha3/sha3module.o |
| 134 | +
|
| 135 | + [stderr] |
| 136 | + error: command 'gcc' failed: No such file or directory |
| 137 | +``` |
| 138 | + |
| 139 | +Then, you'll need to install it with your system package manager, e.g., to resolve the error above: |
| 140 | + |
| 141 | +```console |
| 142 | +$ apt-get install gcc |
| 143 | +``` |
| 144 | + |
| 145 | +!!! note |
| 146 | + |
| 147 | + When using the uv-managed Python versions, it's common to need `clang` installed instead of |
| 148 | + `gcc`. |
| 149 | + |
| 150 | +### Header or library is missing |
| 151 | + |
| 152 | +If the build error mentions a missing header or library, e.g., a `.h` file, then you'll need to |
| 153 | +install it with your system package manager. |
| 154 | + |
| 155 | +For example, installing `pygraphviz` requires Graphviz to be installed: |
| 156 | + |
| 157 | +<!-- docker run --platform linux/x86_64 -it ghcr.io/astral-sh/uv:python3.12-bookworm /bin/bash -c "uv pip install --system 'pygraphviz'" --> |
| 158 | + |
| 159 | +```hl_lines="18-19" |
| 160 | +× Failed to download and build `pygraphviz==1.14` |
| 161 | +╰─▶ Build backend failed to build wheel through `build_wheel` (exit status: 1) |
| 162 | +
|
| 163 | + [stdout] |
| 164 | + running bdist_wheel |
| 165 | + running build |
| 166 | + running build_py |
| 167 | + ... |
| 168 | + gcc -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O3 -Wall -fPIC -DSWIG_PYTHON_STRICT_BYTE_CHAR -I/root/.cache/uv/builds-v0/.tmpgLYPe0/include -I/usr/local/include/python3.12 -c pygraphviz/graphviz_wrap.c -o |
| 169 | + build/temp.linux-x86_64-cpython-312/pygraphviz/graphviz_wrap.o |
| 170 | +
|
| 171 | + [stderr] |
| 172 | + ... |
| 173 | + pygraphviz/graphviz_wrap.c:9: warning: "SWIG_PYTHON_STRICT_BYTE_CHAR" redefined |
| 174 | + 9 | #define SWIG_PYTHON_STRICT_BYTE_CHAR |
| 175 | + | |
| 176 | + <command-line>: note: this is the location of the previous definition |
| 177 | + pygraphviz/graphviz_wrap.c:3023:10: fatal error: graphviz/cgraph.h: No such file or directory |
| 178 | + 3023 | #include "graphviz/cgraph.h" |
| 179 | + | ^~~~~~~~~~~~~~~~~~~ |
| 180 | + compilation terminated. |
| 181 | + error: command '/usr/bin/gcc' failed with exit code 1 |
| 182 | +
|
| 183 | + hint: This error likely indicates that you need to install a library that provides "graphviz/cgraph.h" for `[email protected]` |
| 184 | +``` |
| 185 | + |
| 186 | +To resolve this error on Debian, you'd install the `libgraphviz-dev` package: |
| 187 | + |
| 188 | +```console |
| 189 | +$ apt install libgraphviz-dev |
| 190 | +``` |
| 191 | + |
| 192 | +Note that installing the `graphviz` package is not sufficient, the development headers need to be |
| 193 | +installed. |
| 194 | + |
| 195 | +!!! tip |
| 196 | + |
| 197 | + To resolve an error where `Python.h` is missing, install the [`python3-dev` package](https://packages.debian.org/bookworm/python3-dev). |
| 198 | + |
| 199 | +### Module is missing or cannot be imported |
| 200 | + |
| 201 | +If the build error mentions a failing import, consider |
| 202 | +[disabling build isolation](../concepts/projects/config/#build-isolation). |
| 203 | + |
| 204 | +For example, some packages assume that `pip` is available without declaring it as a build |
| 205 | +dependency: |
| 206 | + |
| 207 | +<!-- docker run --platform linux/x86_64 -it ghcr.io/astral-sh/uv:python3.12-bookworm-slim /bin/bash -c "uv pip install --system chumpy" --> |
| 208 | + |
| 209 | +```hl_lines="7" |
| 210 | +× Failed to download and build `chumpy==0.70` |
| 211 | +╰─▶ Build backend failed to determine requirements with `build_wheel()` (exit status: 1) |
| 212 | +
|
| 213 | + [stderr] |
| 214 | + Traceback (most recent call last): |
| 215 | + File "<string>", line 9, in <module> |
| 216 | + ModuleNotFoundError: No module named 'pip' |
| 217 | +
|
| 218 | + During handling of the above exception, another exception occurred: |
| 219 | +
|
| 220 | + Traceback (most recent call last): |
| 221 | + File "<string>", line 14, in <module> |
| 222 | + File "/root/.cache/uv/builds-v0/.tmpvvHaxI/lib/python3.12/site-packages/setuptools/build_meta.py", line 334, in get_requires_for_build_wheel |
| 223 | + return self._get_build_requires(config_settings, requirements=[]) |
| 224 | + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ |
| 225 | + File "/root/.cache/uv/builds-v0/.tmpvvHaxI/lib/python3.12/site-packages/setuptools/build_meta.py", line 304, in _get_build_requires |
| 226 | + self.run_setup() |
| 227 | + File "/root/.cache/uv/builds-v0/.tmpvvHaxI/lib/python3.12/site-packages/setuptools/build_meta.py", line 522, in run_setup |
| 228 | + super().run_setup(setup_script=setup_script) |
| 229 | + File "/root/.cache/uv/builds-v0/.tmpvvHaxI/lib/python3.12/site-packages/setuptools/build_meta.py", line 320, in run_setup |
| 230 | + exec(code, locals()) |
| 231 | + File "<string>", line 11, in <module> |
| 232 | + ModuleNotFoundError: No module named 'pip' |
| 233 | +``` |
| 234 | + |
| 235 | +To resolve this error, pre-install the build dependencies then disable build isolation for the |
| 236 | +package: |
| 237 | + |
| 238 | +```console |
| 239 | +$ uv pip install pip setuptools |
| 240 | +$ uv pip install chumpy --no-build-isolation-package chumpy |
| 241 | +``` |
| 242 | + |
| 243 | +Note you will need to install the missing package, e.g., `pip`, _and_ all the other build |
| 244 | +dependencies of the package, e.g, `setuptools`. |
| 245 | + |
| 246 | +### Old version of the package is built |
| 247 | + |
| 248 | +If a package fails to build during resolution and the version that failed to build is older than the |
| 249 | +version you want to use, try adding a [constraint](./settings/#constraint-dependencies) with a lower |
| 250 | +bound (e.g. `numpy>=1.17`). Sometimes, due to algorithmic limitations, the uv resolver tries to find |
| 251 | +a fitting version using unreasonably old packages, which can be prevented by using lower bounds. |
| 252 | + |
| 253 | +For example, when resolving the following dependencies on Python 3.10, uv attempts to build an old |
| 254 | +version of `apache-beam`. |
| 255 | + |
| 256 | +```title="requirements.txt" |
| 257 | +dill<0.3.9,>=0.2.2 |
| 258 | +apache-beam<=2.49.0 |
| 259 | +``` |
| 260 | + |
| 261 | +<!-- docker run --platform linux/x86_64 -it ghcr.io/astral-sh/uv:python3.10-bookworm-slim /bin/bash -c "printf 'dill<0.3.9,>=0.2.2\napache-beam<=2.49.0' | uv pip compile -" --> |
| 262 | + |
| 263 | +```hl_lines="1" |
| 264 | +× Failed to download and build `apache-beam==2.0.0` |
| 265 | +╰─▶ Build backend failed to determine requirements with `build_wheel()` (exit status: 1) |
| 266 | +
|
| 267 | + [stderr] |
| 268 | + ... |
| 269 | +``` |
| 270 | + |
| 271 | +Adding a lower bound constraint, e.g., `apache-beam<=2.49.0,>2.30.0`, resolves this build failure as |
| 272 | +uv will avoid using an old version of `apache-beam`. |
| 273 | + |
| 274 | +Constraints can also be defined for indirect dependencies using `constraints.txt` files or the |
| 275 | +[`constraint-dependencies`](./settings.md#constraint-dependencies) setting. |
| 276 | + |
| 277 | +### Package is only needed for an unused platform |
| 278 | + |
| 279 | +If locking fails due to building a package from a platform you do not need to support, consider |
| 280 | +[limiting resolution](../concepts/projects/config/#limited-resolution-environments) to your |
| 281 | +supported platforms. |
| 282 | + |
| 283 | +### Package does not support all Python versions |
| 284 | + |
| 285 | +If you support a large range of Python versions, consider using markers to use older versions for |
| 286 | +older Python versions and newer versions for newer Python version. For example, `numpy` only |
| 287 | +supports four Python minor version at a time, so to support a wider range of Python versions, e.g., |
| 288 | +Python 3.8 to 3.13, the `numpy` requirement needs to be split: |
| 289 | + |
| 290 | +``` |
| 291 | +numpy>=1.23; python_version >= "3.10" |
| 292 | +numpy<1.23; python_version < "3.10" |
| 293 | +``` |
| 294 | + |
| 295 | +### Package is only usable on a specific platform |
| 296 | + |
| 297 | +If locking fails due to building a package that is only usable on another platform, you can |
| 298 | +[provide dependency metadata manually](https://docs.astral.sh/uv/reference/settings/#dependency-metadata) |
| 299 | +to skip the build. uv can not verify this information, so it is important to specify correct |
| 300 | +metadata when using this override. |
0 commit comments