From: Jeff King Date: Mon, 30 Mar 2026 22:33:46 GMT Subject: Re: [PATCH 01/16] midx-write: handle noop writes when converting incremental chains Message-ID: <20260330223346.GA41843@coredump.intra.peff.net> In-Reply-To: <0c7a68ca5a777cce9c0c951643ea2220b1b56a8c.1774820449.git.me@ttaylorr.com> On Sun, Mar 29, 2026 at 05:40:53PM -0400, Taylor Blau wrote: > This comparison fails when we have an incremental MIDX chain with > at least two layers, since we do not recursively peel through earlier > layers, instead treating the `->pack_names` array of the tip MIDX layer > as containing all `m->num_packs + m->num_packs_in_base` packs. > > Adjust this to instead look through the MIDX layers one by one when > comparing pack names. While we're at it, fix a typo above in the same > function. Makes sense. > @@ -1208,14 +1208,16 @@ static bool midx_needs_update(struct multi_pack_index *midx, struct write_midx_c > BUG("same pack added twice?"); > } > > - for (uint32_t i = 0; i < ctx->nr; i++) { > - strbuf_reset(&buf); > - strbuf_addstr(&buf, midx->pack_names[i]); > - strbuf_strip_suffix(&buf, ".idx"); > + for (struct multi_pack_index *m = midx; m; m = m->base_midx) { > + for (uint32_t i = 0; i < m->num_packs; i++) { > + strbuf_reset(&buf); > + strbuf_addstr(&buf, m->pack_names[i]); > + strbuf_strip_suffix(&buf, ".idx"); > > - if (!strset_contains(&packs, buf.buf)) > - goto out; > - strset_remove(&packs, buf.buf); > + if (!strset_contains(&packs, buf.buf)) > + goto out; > + strset_remove(&packs, buf.buf); > + } I wondered here if we would also need to adjust the check of the count, but it is already OK. Just above the context is: if (ctx->nr != midx->num_packs + midx->num_packs_in_base) goto out; which handles the layered midx. Since we were counting up to ctx->nr before, which includes packs in base midx layers, I guess these old lines: > - for (uint32_t i = 0; i < ctx->nr; i++) { > [...] > - strbuf_addstr(&buf, midx->pack_names[i]); was probably reading garbage off the end of the array. Now we iterate through each layer individually. We assume that the total count matches ctx->nr, otherwise the midx struct is totally corrupt. We could count them and double-check, I guess, but I don't think there's much point (it would be a BUG() if they were mismatched). So looks good to me. -Peff