Ports using compiler_name_map are missing from the portindex

[yves.dubief@…]

Hi I am trying to install gcc45 and gcc47 without success. sudo port install gcc45 (or 47 or 48) returns Port gcc45 not found. The ports are also absent from the Available ports on ​http://www.macports.org/ports.php Is this a database problem? It seems to be linked with a recent update as I was able to compile gcc47 on a different mac a couple of weeks ago.

[neverpanic (Clemens Lang)]

What version of MacPorts do you run? From what I see, these ports need MacPorts >= 2.1.0.

Can you run

cd $(port dir mplayer2)/../../lang/gcc48 && port info

and paste the output?

[raimue (Rainer Müller)]

The PortIndex on the rsync server is currently incomplete and missing for example the gcc45 port.

rsync -rv rsync://rsync.macports.org/release/tarballs/PortIndex* /tmp/
grep gcc45 /tmp/PortIndex_darwin_*/PortIndex.quick

This does not return any hits, although I would have expected an entry in each PortIndex.quick.

[mf2k (Frank Schima)]

I see this problem on regular non-trunk installs of Macports 2.1.2. If I manually run a portindex, it adds 36 ports including the missing gcc ports.

[jmroot (Joshua Root)]

The problem seems to have come up only in the last day, as a machine I synced with the pregenerated PortIndex yesterday has no problem looking up the gcc ports.

[jmroot (Joshua Root)]

Looks like the culprit is r102108 and the other similar commits. The compiler_name_map isn't exactly meant to be a public API in the first place, which is why it's semi-hidden inside the portconfigure namespace, but the issue here is that there's no check for whether it actually contains a mapping for the current configure.compiler value. The server generating the indices runs Linux, so the Xcode compilers are not available. (This also means you can't rely on having configure.compiler set "correctly" at index time.)

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #37824.

[jeremyhu (Jeremy Huddleston Sequoia)]

Ok, so then what is the alternative? Can we just get a fix for #32542, so we can avoid having to hack around this? What's a safe thing to do in the interim until #32542 is actually fixed?

[jmroot (Joshua Root)]

I guess you'd a check like

if {[info exists portconfigure::compiler_name_map(${configure.compiler})]} {

instead of calling compiler_is_port, as the immediate workaround. That may be what compiler_is_port should just do anyway, come to think of it. If we decide to make this stuff public, it just needs to be a little more robust.

[jeremyhu (Jeremy Huddleston Sequoia)]

r102174

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #37850.