On Mon Feb 10, 2025 at 10:37 PM CET, Harald van Dijk wrote: > On 10/02/2025 21:15, Sertonix wrote: >> On Mon Feb 10, 2025 at 10:08 PM CET, Harald van Dijk wrote: >>> Hi, >>> >>> On 10/02/2025 19:14, Sertonix wrote: >>>> The test-test example in the comment wasn't actually fixed before as >>>> one can see by running >>>> env -i "test-test=test" busybox ash -c 'export -p' >>>> >>>> This previously incorrectly resulted in >>>> export test >>> >>> You're right that that output is incorrect. However... >>> >>>> But now it correctly results in >>>> export 'test-test=test' >>> >>> ...this is not the correct result either. As the comment in the code >>> says, the output is meant to be usable for 'eval'. This output is not >>> usable for 'eval' and would result in a "bad variable name" error. >> >> This is not the case. I changed the variable name checking to be more >> relaxed for export so it works as expected. You can try this to verify: >> env -i "test-test=test" busybox ash -c 'eval "$(export -p)"' > > Oh, apologies, I missed that, I tested the output in an unpatched > busybox instead. I did not see in your message that you changed that as > well. > > That change still looks wrong: unless there is some special permission > that I am not seeing, 'export test-test=test' is supposed to either > result in an error, or in the specified variable being assigned in the > way that POSIX specifies variable assignments work. The way POSIX > specifies variable assignments work mean the variable behaves like any > other variable. The syntax might not always allow that variable to be > referenced, but consider > > env -i ./busybox ash -c 'export foo=bar baz-=quux; export -p; set' > > *If* the assignment of baz- succeeds, then the requirements for 'export' > say that it becomes a shell variable, and the requirements for 'set' say > it gets printed there as well. That does not happen with your patch.
The POSIX doc for set (and only for set) states that "the output shall be suitable for reinput to the shell, setting or resetting, as far as possible". The way I read this is that skipping a variable name that can't be parsed again would be ok. > I believe my below comment still applies, the variable needs to be > skipped when printing instead. > >>> The variable needs to be skipped when printing instead. >>> >>> Cheers, >>> Harald van Dijk _______________________________________________ busybox mailing list [email protected] https://lists.busybox.net/mailman/listinfo/busybox
