2026-10-26T04:54:09Z https://nostr.now wicked witch of the unix https://nostr.now/npub1ku39qhads3p8a5tgr5krrwl30239z00v8xzqvnlxjtv6znmewwesr687t7 https://social.vlhl.dev/media/d0e5b2ebcfa8f0baf2d873e370ae22d56519931bdb9cd5c255c6af3ea843b730.jpg https://social.vlhl.dev/media/d0e5b2ebcfa8f0baf2d873e370ae22d56519931bdb9cd5c255c6af3ea843b730.jpg https://nostr.now/nevent1qqsvr57w357w9kep48m5yc557s4kuqf4swj43magw7dq8fgl6tegm5szyzmjy5zl4kzyylk3dqwjcvdm79a2y5faasucgpj0u6fdng2009emx4kna44 In reply to <a href='/nevent1qqsy6nhukfyv26rcqe5udhjmjw2avklyfktpg02tt0m47qh8e0tdzusg683th'>nevent1q…83th</a><br/>_________________________<br/><br/>as in like: <br/> <br/>``` <br/>type foo; <br/>foo.nya = ...; <br/>foo.mew = ...; <br/>defer { <br/> free(foo.nya); <br/> free(foo.mew); <br/>} <br/>``` <br/> <br/>or? i&#39;m not sure what &#34;more local cleanup&#34; would mean <br/> <br/>i&#39;m going by the most common api pattern i see out there, that is, if `make_foo()` returns success, it means the struct was fully allocated, and free_foo() deallocates all it needs <br/> <br/>if make_foo() can fail, then i&#39;d be something more like: <br/> <br/>```c <br/>type foo; <br/>if ((err = make_foo(&amp;foo))) <br/> return err; <br/>defer free_foo(foo); <br/>...; <br/>type nya; <br/>if ((err = make_nya(&amp;nya, &amp;foo))) <br/> return err; <br/>defer free_nya(nya); <br/>``` <br/> <br/>(defer could also help inside those constructors, by a pattern like <br/> <br/>```c <br/>enum myerr err = OK; <br/>struct foo *foo = ...; <br/> <br/>defer if (err != OK) free_foo(foo); <br/>...; <br/>if (some_cond() == FAIL) <br/> return (err = COND_FAILED); <br/>``` <br/> <br/>ofc this also depends on your data structure for foo <br/> <br/>as a side note, for &#34;handle every kind of failure&#34;: <br/> <br/>it&#39;s a lot better if you employ some sort of zii (zero is initialization), since then you can calloc() foo, (or for a stack structure, `foo = {};`) to zero it out; have free_foo cleanup when non-0 where possible, it&#39;s one of the reasons `free(NULL);` is a noop <br/> <br/>i don&#39;t have to care much *where* something failed and the destructor got called, since everything that wasn&#39;t allocated would be a zero or some other value that can either be checked or noops (the only exception is usually fds where you have to assign -1, for this you can use a `foo = FOO_INIT;` macro, or just assign by hand, `foo = { .myfd = -1 }`, it&#39;s annoying but works) 2026-10-11T09:42:53Z https://nostr.now/nevent1qqswpqmfkrmdq0e88h0lah9qh50ltw40n2975tka9ec83ewsfrd7hmgzyzmjy5zl4kzyylk3dqwjcvdm79a2y5faasucgpj0u6fdng2009emxlrnedg In reply to <a href='/nevent1qqsglawsu83y3tw5l300296nhdm4rgf4s52nc9js3tm6cqpf6rjc2xc68ewvu'>nevent1q…ewvu</a><br/>_________________________<br/><br/>i don&#39;t see how that constrains anything? deferred blocks would be called on basically the same order as c++ destructors, plus you can defer whole blocks if you need more at cleanup time <br/> <br/>and i take <br/> <br/>```c <br/>type *foo = make_foo(); <br/>defer free_foo(foo); <br/>// ...; <br/>if ((err = some_call())) <br/> return err; <br/>type *nya = make_nya(&amp;foo); <br/>defer free_nya(nya); <br/> <br/>if ((err = some_other_call())) <br/> return err; <br/>//...; <br/>``` <br/> <br/>over <br/> <br/>```c <br/>type *foo = make_foo(); <br/>// ...; <br/>if ((err = some_call())) <br/> goto cleanup_foo; <br/>// ...; <br/>type *nya = make_nya(&amp;foo); <br/> <br/>if ((err = some_other_call())) <br/> goto cleanup_nya; <br/>// ...; <br/>cleanup_nya: <br/> free_nya(nya); <br/>cleanup_foo: <br/> free_foo(foo); <br/> return err; <br/>``` <br/> <br/>for the defer example, i see foo, i see deferring it&#39;s cleanup, i don&#39;t think about foo&#39;s resources on exit anymore <br/> <br/>on the goto example, i constantly have to think about it, if i change some calls around i gotta double check which labels it&#39;s jumping to, it&#39;s a lot easier to make a mistake refactoring and get a double-free or leak <br/> <br/>and to me, defer is not any less explicit, for the goto example if i&#39;m looking at the cleanup at the end i have to be aware of every single point in the function that jumps to it, i have to personally keep track of allocation order to make sure cleanup order is correct (nya takes a foo, so foo *has* to be cleaned up later) <br/> <br/>defer just handles all that by design, all i have to do is keep the defer close to the allocation and now i don&#39;t have to worry much about ordering on the 99% of common cases; it&#39;s still explicit, it&#39;s still there in the function telling me the flow, it&#39;s not like destructors living in another file completely and being called implicitly <br/> <br/>(and sure defer isn&#39;t idea on every case, it&#39;s made to avoid common mistakes, just like we have `for (init; cond; end) body;` loops instead of `init; again: if (cond) goto over; body; end; goto again; over:`) 2026-10-11T09:23:15Z https://nostr.now/nevent1qqsgkr94q372g2qsckqd0ekzymzf3plvg2qly09a2wckmy76jzmg78qzyzmjy5zl4kzyylk3dqwjcvdm79a2y5faasucgpj0u6fdng2009emx5tkcs7 In reply to <a href='/nevent1qqsx7w4ey7jrrk34ny92nvgznrce7qqp35k5k3hrsxewelyynlc3psswdzx2k'>nevent1q…zx2k</a><br/>_________________________<br/><br/>reverse order when scope ends, it&#39;s not much more confusing than the last expression in a for loop, imo<br/><br/>(i also find it often clear than a sequence of goto cleanup_N; when you&#39;re dealing with more than one exit point) 2026-10-11T08:47:24Z https://nostr.now/nevent1qqsfaqu566s7smlzgn2kxm4uvaxe3d658ex00fqk4rdddteym6zayuczyzmjy5zl4kzyylk3dqwjcvdm79a2y5faasucgpj0u6fdng2009emxlwe904 when&#39;s this whole &#34;c is portable assembly&#34; thing gonna end<br/><br/>it&#39;s just not 2026-10-11T08:39:35Z https://nostr.now/nevent1qqsx7f9ul9n3q4ggnlly8vdh7tj423849h7388gjac2nha3cktka6sgzyzmjy5zl4kzyylk3dqwjcvdm79a2y5faasucgpj0u6fdng2009emx3snkqf In reply to <a href='/nevent1qqsfvherhy6s4vge8vkm9zqmkug44x02qduafrdlzduwm55ggrdvumqpwdjud'>nevent1q…djud</a><br/>_________________________<br/><br/>only valve and nintendo are trying new things with controllers, tbh -- and valve is hitting it spot on with gyro and dual touchpads<br/><br/>i&#39;m not a fan of a law requiring a specific controller layout at all, everything would become an xbox controller<br/><br/>if we have to have it, then i&#39;d rather it be at most require adaptive icons in the ui like how steam does it, changing the displayed layout based on the controller being used, and nothing else 2026-05-19T00:06:05Z