IRC channel logs

<oriansj>stikonas: thank you. I guess I was too tired to notice that.

<stikonas>well, this happens if you work too late

<stikonas>happened to me too...

<stikonas>anyway, I've got the next interesting thing

<stikonas>fossy, oriansj: do you want conditionals in kaem?

<oriansj>well with it we could enable the use of kaem scripts for mescc-tools and M2-Planet tests.

<stikonas>so I mostly got it working

<oriansj>nice

<stikonas>it's not 100% nice with output in --verbose mode but not too bad

<oriansj>So it'll be some work

<stikonas>well, let me paste how it looks

<oriansj>and I have some work to do on untar to address: https://www.openwall.com/lists/oss-security/2021/10/03/1

<stikonas>so for this kame file https://paste.debian.net/1214547/

<stikonas>I get the following output https://paste.debian.net/1214548/

<oriansj>so didnt' echo the else

<stikonas>if I have else block, it echos just else but nothing after else

<stikonas>but else is printed

<stikonas>and then in that case fi is not printed

<stikonas>so maybe it's acceptable...

<stikonas>and this is the implementation: https://github.com/stikonas/mescc-tools/commit/8831dac1366c979df3e1d226e30bf6c8f5729c46

<stikonas>ok, I made a PR now

<oriansj>the old_VERBOSE might not be the right way.

<stikonas>fossy: you might want to take a look https://github.com/oriansj/mescc-tools/pull/28

<stikonas>hmm

<stikonas>oriansj: what do you suggest?

<stikonas>well, even with this PR conditionals are not fully useful as one needs test command...

<stikonas>but one step at a time...

<oriansj>break it into 2 loops, one for the if and one for the else (if it exists)

<stikonas>yeah, maybe that would be clearer...

<stikonas>hmm, somehow 2 loops look worse...

<stikonas>or at least my implementation

<stikonas>wrote something like https://paste.debian.net/1214551/ but I don't like it...

<stikonas>it's basically twice as much code

<stikonas>probably can't do much more with kaem without requiring extra syscalls

<stikonas>globs need opendir, $() needs dup2. for loops might not need extra syscalls but tricky to implement with current design (kaem reads line by line)

<oriansj>stikonas: absolutely true. kaem was never designed for half the things it does. They kind of grew organically.

<oriansj>So the question becomes should we continue to extend kaem or should we pick up a better shell?

<stikonas>well, we have bash

<stikonas>which is built fairly quickly

<stikonas>but I don't know if we want something between bash and kaem

<stikonas>at some point we might get gash

<oriansj>for example: https://github.com/rain-1/s

<stikonas>but we should definitely keep a shell (kaem) that requires minimal number of syscalls

<stikonas>I think I saw this s

<stikonas>I don't remember what was the issue with it that prevented it from being included into live-bootstrap

<oriansj>it wasn't buildable previously as if I remember correctly.

<stikonas>possibly

<stikonas>maybe tcc had problems with it

<stikonas>or maybe uses more libc than mes has

<stikonas>probably the latter

<oriansj>but that is the problem with shell interpreters.

<oriansj>They are very heavy in syscalls required and can easily turn into unmaintainable messes

<stikonas>well, I started working on if/else cause that might be useful for multiple arch support...

<oriansj>with get_machine I am guessing

<stikonas>hmm

<stikonas>get_machine needs modifications I guess

<oriansj>no

<stikonas>we can't redirect stdout

<oriansj>but I got an idea

<stikonas>hmm

<stikonas>well, it can return arch number...

<stikonas>oh, that's still trickyu

<stikonas>hmm

<oriansj>get_machine -if x86 -exec "./kaem -f foo"

<stikonas>do we actually need get_machine?

<stikonas>if we just define ARCH="whatever" in initial kaem.arch

<oriansj>no we could create something custom

<stikonas>although, initial kaem.arch is kaem-minimal

<stikonas>so can't define much

<stikonas>well, actually we can define after we build full kaem

<oriansj>but kaem might benefit from being able to do: ./kaem -f foo.$(get_machine)

<stikonas>but $() needs extra syscalls

<stikonas>dup2

<oriansj>yeah unfortunately

<stikonas>hence I'm thinking of just defining it as env variable

<stikonas>which should be doable

<stikonas>(possibly just before after.kaem)

<stikonas>or maybe even earlier

<oriansj>well extending kaem to have access to the uname data would be trivial

<fossy>stikonas[m]: conditionals would be nice it was ob my todo list a while ago

<pabs3>will bootstrappable be at LibrePlanet? https://www.fsf.org/blogs/community/libreplanet-2022-returns-online-submit-your-session-proposal-by-dec-1

<oriansj>pabs3: well the last 3 proposals were rejected. So I think I need to make a much better proposal to get accepted.

<stikonas>fossy: if you have some spare time maybe review my kaem PR?

IRC channel logs

2021-10-07.log