legacy-support: can't read "ls_cache_incpath": no such variable

[ryandesign (Ryan Carsten Schmidt)]

While mirroring mercurial, this error occurred, which seems to relate to legacy support:

​https://build.macports.org/builders/jobs-mirror/builds/327006/steps/mirror/logs/stdio

Error: mportopen file:///opt/local/var/buildworker/jobs/jobs-mirror/build/ports/devel/mercurial failed: can't read "ls_cache_incpath": no such variable

[cjones051073 (Chris Jones)]

Idon't see anything wrong with what the legacysupport PG is doing with this variable. It must relate to something the port is doing ?

[jmroot (Joshua Root)]

Mercurial doesn't use legacysupport directly, but via cargo_fetch.

[ryandesign (Ryan Carsten Schmidt)]

Right, with the +rust variant.

This is not a new problem nor intermittent; I see it in every mirroring log for mercurial I've looked at.

[jmroot (Joshua Root)]

Only affects older OS versions apparently (presumably ones on which legacysupport is active). Debug output:

% port -d info mercurial +rust
DEBUG: Changing to port directory: /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/devel/mercurial
DEBUG: OS darwin/10.8.0 (macOS 10.6.8) arch i386
DEBUG: Sourcing PortGroup compiler_wrapper 1.0 from /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/_resources/port1.0/group/compiler_wrapper-1.0.tcl
DEBUG: Re-registering default for distname
DEBUG: Re-registering default for build.cmd
DEBUG: Re-registering default for build.target
DEBUG: Re-registering default for destroot.cmd
DEBUG: Re-registering default for destroot.destdir
DEBUG: Re-registering default for destroot.target
DEBUG: Sourcing PortGroup python 1.0 from /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/_resources/port1.0/group/python-1.0.tcl
DEBUG: adding the default universal variant
DEBUG: Reading variant descriptions from /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/_resources/port1.0/variant_descriptions.conf
DEBUG: Executing variant rust provides rust
DEBUG: Sourcing PortGroup compiler_blacklist_versions 1.0 from /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/_resources/port1.0/group/compiler_blacklist_versions-1.0.tcl
DEBUG: Sourcing PortGroup legacysupport 1.1 from /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/_resources/port1.0/group/legacysupport-1.1.tcl
DEBUG: Re-registering default for dist_subdir
DEBUG: Re-registering default for extract.only
DEBUG: Sourcing PortGroup muniversal 1.0 from /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/_resources/port1.0/group/muniversal-1.0.tcl
DEBUG: Sourcing PortGroup cargo_fetch 1.0 from /opt/local/var/macports/sources/rsync.macports.org/macports/release/tarballs/ports/_resources/port1.0/group/cargo_fetch-1.0.tcl
DEBUG: Running callback portconfigure::add_automatic_compiler_dependencies
DEBUG: Chosen compiler macports-clang-11 is provided by a port, adding dependency
DEBUG: Adding depends_build port:clang-11
DEBUG: Adding depends_lib libcxx
DEBUG: Adding depends_skip_archcheck clang-11
DEBUG: Finished running callback portconfigure::add_automatic_compiler_dependencies
DEBUG: Running callback portbuild::add_automatic_buildsystem_dependencies
DEBUG: Finished running callback portbuild::add_automatic_buildsystem_dependencies
DEBUG: Running callback portstartupitem::add_notes
DEBUG: Finished running callback portstartupitem::add_notes
DEBUG: Running callback compwrap::configure_envs
DEBUG: Finished running callback compwrap::configure_envs
DEBUG: Running callback python_add_dependencies
DEBUG: Finished running callback python_add_dependencies
DEBUG: Running callback legacysupport::add_legacysupport
DEBUG: legacysupport: Will append path:lib/libMacportsLegacySupport.dylib:legacy-support to depends_lib
DEBUG: can't read "ls_cache_incpath": no such variable
    while executing
"if { ${ls_cache_incpath} ne "" } {
            foreach f ${ls_cache_ldflags}  { configure.ldflags-delete ${f} }
            foreach f ${ls_cache_cppfl..."
    (procedure "legacysupport::add_legacysupport" line 11)
    invoked from within
"${callback}"
    (procedure "port::run_callbacks" line 5)
    invoked from within
"port::run_callbacks"
    invoked from within
"$workername eval {port::run_callbacks}"
    (procedure "mportopen" line 66)
    invoked from within
"mportopen $porturl [array get options] [array get merged_variations]"
Error: Unable to open port: can't read "ls_cache_incpath": no such variable

[ryandesign (Ryan Carsten Schmidt)]

Yes, there we go. I can reproduce it on High Sierra with:

port info mercurial +rust legacysupport.newest_darwin_requires_legacy=99

(The machine doing the mirroring still runs El Capitan.)

[jmroot (Joshua Root)]

So yeah, whenever the callback actually does something, it breaks. I would guess it's because mercurial is including a portgroup inside a variant, which means some things that the portgroup code thinks will be run in a global context are actually run in a proc context. There's a reason why PortGroup is only supposed to be used at the top of the Portfile.

[ryandesign (Ryan Carsten Schmidt)]

But mercurial only needs cargo_fetch in the rust variant. The cargo_fetch portgroup does a bunch things immediately when it's included. Will that mess up the mercurial port when the rust variant is not selected?

[jmroot (Joshua Root)]

Probably yeah. It's not a great design. Workarounds could be to include the portgroup conditionally with variant_isset, or maybe somehow make legacysupport work regardless of the context it runs in?

[cjones051073 (Chris Jones)]

If anyone has suggestions what to do make the cargo_fetch/legacysuppoirt PG work better when included inside a variant I am open to suggestions. I guess the 'problem' lines are the global variable caches at

​https://github.com/macports/macports-ports/blob/da89704a46ac2b5a9771b8264e3876e4d46199e9/_resources/port1.0/group/legacysupport-1.1.tcl#L150

if anyone ha suggestions what to do with these instead, I am open to suggestions.

To me though the simpliest solution is to avoid including PGs in this way, in which case Joshua's suggestion of

if {[variant_isset rust]} {
    PortGroup cargo_fetch 1.0
}

at the top of the mercurial Portfile, seems the best fix.

[Chris Jones <jonesc@…>]

In 1dc522b2a72f49baa0b8116fde7b4a48fa219e77/macports-ports (master):

mercurial: avoid including cargo_fetch PG inside variant
Closes: #63834

[ryandesign (Ryan Carsten Schmidt)]

Thanks, I agree that does work around the problem.