guix: update per-host disk-space estimates in build gate

djkazic commented at 5:21 pm on November 17, 2025: none

Today, the build gates are obsolete. I found this out after seeing someone flag this on a commit setting -static-libgcc. Note that the Linux builds came in at the expected amount of disk space utilization, so I did not touch the values for them.

This PR aims to readjust the build gate disk space per-host estimates for darwin and mingw based on some testing I did locally in my build VM.

Details: (Copied from https://github.com/kevkevinpal/bitcoin/issues/200):

I ran HOSTS="x86_64-w64-mingw32" ./contrib/guix/guix-build

After (+10 GiB):

0Filesystem     1G-blocks  Used Available Use% Mounted on
1/dev/sda5           512G   98G      393G  20% /gnu/store

These numbers makes this change seem reasonable (1.5x safety factor):

0-        *mingw*)  required_KiB=7600000 ;;
1+        *mingw*)  required_KiB=15728640 ;;

Then I ran HOSTS="aarch64-linux-gnu" ./contrib/guix/guix-build.

After (+6 GiB):

0Filesystem      Size  Used Avail Use% Mounted on
1/dev/sda5       512G  104G  387G  22% /gnu/store

Then I setup the SDK for macOS. This established a new before snapshot of disk use:

0Filesystem      Size  Used Avail Use% Mounted on
1/dev/sda5       512G  116G  375G  24% /gnu/store

Then I ran HOSTS="x86_64-apple-darwin" ./contrib/guix/guix-build.

After (+5 GiB):

0Filesystem      Size  Used Avail Use% Mounted on
1/dev/sda5       512G  121G  370G  25% /gnu/store

IMO, this is grounds for updating this (5 GiB * 1.5x safety factor):

0-        *darwin*) required_KiB=440000 ;;
1+        *darwin*) required_KiB=8388608 ;;

guix: update per-host disk-space estimates in build gate cbddfbac82

DrahtBot added the label Build system on Nov 17, 2025

DrahtBot commented at 5:21 pm on November 17, 2025: contributor

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Code Coverage & Benchmarks

For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/33889.

Reviews

See the guideline for information on the review process.

Type	Reviewers
ACK	bensig

If your review is incorrectly listed, please copy-paste <!–meta-tag:bot-skip–> into the comment that the bot should ignore.

Conflicts

Reviewers, this pull request conflicts with the following ones:

#34638 (guix: Update storage requirement for HOSTS and added warning for Guix overhead by janb84)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

in contrib/guix/guix-build:157 in cbddfbac82

153@@ -154,8 +154,8 @@ avail_KiB="$(df -Pk "$VERSION_BASE" | sed 1d | tr -s ' ' | cut -d' ' -f4)"
154 total_required_KiB=0
155 for host in $HOSTS; do
156     case "$host" in
157-        *darwin*) required_KiB=440000 ;;
158-        *mingw*)  required_KiB=7600000 ;;
159+        *darwin*) required_KiB=8388608 ;;

fanquake commented at 5:31 pm on November 17, 2025:

due to a change made in the build process (-static-libgcc).

Can you explain a bit more how this is related to -static-libgcc, given you haven’t changed the allowance for any Linux builds, and macos/mingw don’t use -static-libgcc?

djkazic commented at 5:34 pm on November 17, 2025:

Yeah, the details are linked in the issue on kev’s fork but the issue was flagged on a commit where static was set. I tested Linux but the disk space used there was in line with expectations so the PR doesn’t modify estimates for that.

I did notice discrepancies for darwin and mingw though, which I’ve attempted to correct for.

maflcko commented at 8:16 am on November 18, 2025:

I did notice discrepancies for darwin and mingw though, which I’ve attempted to correct for.

That is fine, but you’ll have to properly explain the changes.

https://github.com/kevkevinpal/bitcoin/issues/200 could be an oversight where someone ran the build twice by accident?

janb84 commented at 8:57 am on November 18, 2025:

It was my remark, and I was still researching it. My Guix build is using more GiB than before, but it’s not related to 33181. I noticed it on that PR but a master compile (sans 33181) was also consuming more space than before.

djkazic commented at 1:54 pm on November 18, 2025:

@maflcko I proposed the changes to the build gates by just running a build and checking before/after snapshots of disk usage. Do we need more detailed info like tracking which files the extra disk space use comes from?

djkazic commented at 1:56 pm on November 18, 2025:

I’ve updated my PR description to be more in line with our discussions here.

sedited commented at 12:43 pm on February 9, 2026:

still researching it @janb84 any update?

janb84 commented at 2:53 pm on February 9, 2026:

still researching it

@janb84 any update?

Just picked it up again. Would an issue be a good idea to track progress/ exchange ideas ?

bensig commented at 9:42 pm on January 5, 2026: contributor

ACK cbddfbac82c8cab5e9b303f6c6b417e16b0e2501

Math checks out. Testing methodology in the PR description is thorough.

maflcko commented at 3:41 pm on February 20, 2026: member

Looks like #34638 comes to a different result.

maflcko commented at 12:40 pm on March 5, 2026: member

No activity after 2 weeks. I guess this can be closed?

fanquake commented at 1:48 pm on March 5, 2026: member

Guess we’ll continue in #34638.

fanquake closed this on Mar 5, 2026

guix: update per-host disk-space estimates in build gate #33889

Code Coverage & Benchmarks

Reviews

Conflicts