[2/2] build: fail if explicitly requested driver is unbuildable
Checks
Commit Message
When the user passes a list of desired drivers to build via the
"enable_drivers" option, the expectation is that those drivers should be
part of the build. However, if those drivers have either external or
internal dependencies, they still may be silently disabled, for example:
running "meson setup -Denable_drivers=net/iavf build" will successfully
run, but the iavf net driver will not be configured as "common/iavf" is
missing.
We can fix this by setting a flag to indicate when the drivers are
specified via an enable_drivers flag. However, unlike when erroring out
on missing libs, we don't error out if a driver in unbuildable, unless
the driver name explicitly appears in the "enable_drivers" list. This is
implemented this way to ensure that wildcarding still works. For
example: we still want to allow "meson setup -Denable_drivers=net/*" to
work, configuring only the buildable network drivers. While it's true
that this additional restriction may cause some builds to pass when they
should fail, e.g. if the wildcard refers only to a single driver,
implementing things this way avoids massive amounts of complexity, and
is still an improvement on the status-quo.
Suggested-by: Anatoly Burakov <anatoly.burakov@intel.com>
Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
---
drivers/meson.build | 14 ++++++++++++++
1 file changed, 14 insertions(+)
Comments
> From: Bruce Richardson [mailto:bruce.richardson@intel.com]
> Sent: Friday, 1 September 2023 16.24
>
> When the user passes a list of desired drivers to build via the
> "enable_drivers" option, the expectation is that those drivers should be
> part of the build. However, if those drivers have either external or
> internal dependencies, they still may be silently disabled, for example:
> running "meson setup -Denable_drivers=net/iavf build" will successfully
> run, but the iavf net driver will not be configured as "common/iavf" is
> missing.
>
> We can fix this by setting a flag to indicate when the drivers are
> specified via an enable_drivers flag. However, unlike when erroring out
> on missing libs, we don't error out if a driver in unbuildable, unless
> the driver name explicitly appears in the "enable_drivers" list. This is
> implemented this way to ensure that wildcarding still works. For
> example: we still want to allow "meson setup -Denable_drivers=net/*" to
> work, configuring only the buildable network drivers. While it's true
> that this additional restriction may cause some builds to pass when they
> should fail, e.g. if the wildcard refers only to a single driver,
> implementing things this way avoids massive amounts of complexity, and
> is still an improvement on the status-quo.
>
> Suggested-by: Anatoly Burakov <anatoly.burakov@intel.com>
> Signed-off-by: Bruce Richardson <bruce.richardson@intel.com>
> ---
Acked-by: Morten Brørup <mb@smartsharesystems.com>
ARM Ampere server test fails on this patch are lab infra-failures (I did
some updates on the server yesterday) and they can be ignored.
On Sat, Sep 02, 2023 at 01:24:18PM -0400, Patrick Robb wrote:
> ARM Ampere server test fails on this patch are lab infra-failures (I
> did some updates on the server yesterday) and they can be ignored.
Can a retest be run on these now, to clear the errors in patchwork, or have
the issued been fixed?
@@ -41,7 +41,9 @@ disable_drivers = run_command(list_dir_globs, disable_drivers, check: true).stdo
# add cmdline enabled drivers and meson enabled drivers together
enable_drivers = ',' + get_option('enable_drivers')
enable_drivers = run_command(list_dir_globs, enable_drivers, check: true).stdout().split()
+require_drivers = true
if enable_drivers.length() == 0
+ require_drivers = false
enable_drivers = run_command(list_dir_globs, '*/*', check: true).stdout().split()
endif
@@ -155,6 +157,12 @@ foreach subpath:subdirs
build = false
reason = 'requires IOVA in mbuf (set enable_iova_as_pa option)'
endif
+ # error out if we can't build a driver and that driver was explicitly requested,
+ # i.e. not via wildcard.
+ if not build and require_drivers and get_option('enable_drivers').contains(drv_path)
+ error('Cannot build explicitly requested driver "@0@".\n'.format(drv_path)
+ +'\tReason: ' + reason)
+ endif
# get dependency objs from strings
shared_deps = ext_deps
@@ -171,6 +179,12 @@ foreach subpath:subdirs
endif
message('Disabling @1@ [@2@]: missing internal dependency "@0@"'
.format(d, name, 'drivers/' + drv_path))
+ # error out if we can't build a driver and that driver was explicitly
+ # requested, i.e. not via wildcard.
+ if require_drivers and get_option('enable_drivers').contains(drv_path)
+ error('Cannot build explicitly requested driver "@0@".\n'.format(drv_path)
+ +'\tPlease enable missing dependency "@0@"'.format(d))
+ endif
else
shared_deps += [get_variable('shared_rte_' + d)]
static_deps += [get_variable('static_rte_' + d)]