Lint and Livecheck targets should not require root access

[rhwood@…]

The commands port lint foo and port livecheck foo should not require write access at the root level to anything. Perhaps we should distinguish between commands that make changes that should survive the port operation and those that should not survive the port operation (The results of the fetch, extract, patch, configure, build, destroot, archive, install, activate, etc... commands need to be maintained after port has finished, but this is not true for lint and livecheck)

[afb@…]

They do have a target_runtype of "always", if that is usable somehow... ?

[wsiegrist@…]

The problem with lint at least seems to be the need to write to the state file which is in the build area and requires root privileges to write to. These sort of operations dont affect the state of the port so the file shouldnt be written to, IMO. Sorry if this is already obvious to the devs but I wanted to comment here since I ran into the issue and my ticket was marked a dup of this one.

[raimue (Rainer Müller)]

It needs

target_state ${org.macports.foo} no

for every target which is meant to be run without writing to the state file.

I tried to make this possible, but one of the remaining issues is that livecheck tries to write to ${workpath}/livecheck.TMP. Where should livecheck write else? For example /tmp? Tcl does not seem to provide any access to the UNIX syscall mktemp(3), which provides some unique name for a temporary file usually.

With the attached patch it will work if livecheck writes to some other place. At the moment it fails with an Permission denied error (after r34288).

Please review.

[afb@…]

Looks good to me, added patch as r34513.

[raimue (Rainer Müller)]

Use /tmp for livecheck in r34515.