Re: [PATCH 1/2] cargo.eclass: for EAPI 9, ban cargo_src_install

On Fri, May 29, 2026 at 11:15:53AM -0400, Eli Schwartz wrote:
> On 5/29/26 2:49 AM, Ionen Wolkens wrote:
> 
> > It's either way for consistency, now the eclass does it in two different
> > ways. And this isn't POSIX sh, we intentionally use bashism everywhere
> > because this is bash.
> 
> 
> No we don't.
> 
> We *take advantage of* bash everywhere. There is nothing to take
> advantage of, here. There are no internet points to be had for
> dogmatically using a social misfeature with no semantic distinction, as
> some kind of "loyalty test: do you reject POSIX?"
> 
> [[ or enhanced string manipulation or local or arrays or [or or or]...
> are all things we have no *qualms* using. This is bash, "loyalty test:
> do you avoid unportable bashisms" isn't a valid reason to write worse code.
> 
> Loyalty tests in both directions are "things we don't do".
> 
> And my objection to "==" is that it is longer / harder to read, serves
> no purpose (valid or otherwise), and practically is known to break

In my eyes I only see a variable assignment with a single = and it makes
it feel more confusing to me (worse if a line break were to put that
on its own line).

Albeit I almost never write POSIX sh (when I do I tend to use an actual
posix shell to be safe), and used to seeing == due to C and several
other languages inspired by it. I know that C == bad, but in C
`if (var = value)` actually does assign the variable too and is a
common mistake.

Just to say that I disagree and I think == is better and will continue
to use it, but won't pursue this further (not my eclass).

> people's brains. My objection is *consistent* with the fact that I
> believe bash upstream should make it a verbose warning to use it that
> warns you you should have used "=". Because it's a *bad* feature.
> 
> Objecting to a bad feature is certainly opinionated, but it isn't
> against tree policy on style -- you may also note I religiously use
> "spaces not tabs" in private shell code, but cheerfully obey actual tree
> policy for ::gentoo, *because* it is policy. :)

-- 
ionen

-----BEGIN PGP SIGNATURE-----

iQFPBAABCAA5FiEEx3SLh1HBoPy/yLVYskQGsLCsQzQFAmoZx+4bFIAAAAAABAAO
bWFudTIsMi41KzEuMTIsMiwyAAoJELJEBrCwrEM0MNYH/1V+BTXm4cI1UeP9qBtf
VWJIyHBLnZeYQ+6pWH8k8k1maJIfzouMdqDYLX3cG433OI+mDy2COOe6Kwj+bF16
eYGj2dZxsgqzzG5rw4+tOKVN8OLLc7qqDP0zaGLbqG0K1mkPlUzxAyZgt50vte5g
aZG97w4XG/5Fr8TnT5PhBLW4Jd66hWb2+iNUbYvIcmsjzv8hJ+HSQwo2bGdmLpu2
glzIbHNwfphKOLEZRuepFvoPnYwjrRyhZDPL2EfTLPXBwMb2/1I2r5PpnpotRM64
eYL8tuNm/N35mpr/wjGkLMXrZDD21d169TN+iBlfF2crqiLWWIF58N5S0C3AiF12
jF0=
=BDsX
-----END PGP SIGNATURE-----