8000 Ship bashbug, libtool-next-version, xml2-config and xslt-config (world rebuild) by bryan-minimal · Pull Request #1443 · gominimal/pkgs · GitHub
[go: up one dir, main page]

Skip to content

Ship bashbug, libtool-next-version, xml2-config and xslt-config (world rebuild) - #1443

Merged
bryan-minimal merged 1 commit into
mainfrom
bin-coverage-low
Nov 13, 2026
Merged

bryan-minimal merged 1 commit into
mainfrom
bin-coverage-low

Conversation

@bryan-minimal
Copy link
Copy Markdown
Member

Low-level half of the bin-coverage audit; #1436 was the leaf half. Each of these builds already runs upstream's install step, and these four scripts land in usr/bin. The outputs named the other binaries one by one, so these were silently dropped.

Rebuilds: 575 (a world rebuild; bash is in base). Queue it on its own night, not next to another world rebuild.

Added

package executable evidence (pinned source) interpreter
bash 5.3 bashbug Makefile.in:630 (.made builds it), :918 (install installs it; our install-strip calls install) #!/bin/sh -; the sh is bash itself, which can't be its own runtime dep → allow_missing_interpreter
libtool 2.6.2 libtool-next-version Makefile.am:286 / Makefile.in:762 (bin_SCRIPTS, unconditional) #! /bin/sh, satisfied by the existing bash runtime dep
libxml2 2.15.4 xml2-config CMakeLists.txt:800-801 (install(PROGRAMS …), unconditional on non-Windows) #! /bin/sh, no sh runtime dep → allow_missing_interpreter (as #1436 did for jemalloc.sh)
libxslt 1.1.45 xslt-config CMakeLists.txt:554-555 same as xml2-config

Sources: the gs:// tarballs were sha256-verified against build.ncl. libxml2/libxslt were read from a clone of the pinned tag, because GitHub's archive bytes can't be fetched here to compare the sha.

One behavior note: with xml2-config/xslt-config on PATH, a consumer's configure that prefers them over pkg-config will now find them. For shared builds --libs prints just -lxml2 / -lxslt -lxml2, the same set pkg-config gives, so I don't expect breakage. The world rebuild will show it if there is any.

Checked and left out

package executable why
sqlite 3.54.1 sqldiff, sqlite3_rsync Not in the source we build: the autoconf amalgamation tarball has no tool/ sources and its Makefile.in installs only sqlite3. Shipping them means switching to the full sqlite-src zip, which is a separate change.
cmake 4.4.5 ccmake Built only when Source/Checks/Curses.cmake finds curses + form, and cmake has no ncurses dep. Adding one grows cmake's closure for a TUI; that's a decision, not a dropped file.
tcl 8.6.18 tclsh Upstream installs only tclsh8.6 (unix/Makefile.in:817); the bare tclsh link is a distro convention (LFS/Arch add it).
zstd 1.5.7 pzstd Lives in contrib/pzstd; the top-level and programs/Makefile install targets don't install it.

nickel format run on every edited file.

🤖 Generated with Claude Code

Low-level half of the bin-coverage audit (#1436 was the leaf half). Each
build already runs upstream's install step, which installs these four
scripts into usr/bin; our outputs named the other binaries one by one, so
these were dropped.

- bash: bashbug (Makefile.in .made and install targets). #!/bin/sh, and
  the sh it needs is bash itself, so allow_missing_interpreter.
- libtool: libtool-next-version (Makefile.am bin_SCRIPTS). Its /bin/sh
  comes from the bash runtime dep.
- libxml2: xml2-config (CMakeLists.txt install(PROGRAMS)), and
  libxslt: xslt-config (likewise); both #!/bin/sh with no sh runtime dep,
  so allow_missing_interpreter, as #1436 did for jemalloc.sh.

Checked and left out: sqlite's sqldiff/sqlite3_rsync (not in the autoconf
amalgamation tarball we build), cmake's ccmake (needs curses, which cmake
does not depend on), tcl's bare tclsh (upstream installs only tclsh8.6),
zstd's pzstd (contrib/, not installed by make install).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@bryan-minimal
bryan-minimal requested a review from a team as a code owner November 11, 2026 18:19
@coderabbitai
coderabbitai Bot commented Nov 11, 2026
Copy link
Copy Markdown
Contributor

Warning

Review limit reached

  • Run on-demand review

This review includes 4 billable files and costs up to $1.00.

  • Ask an admin to make reviews automatic

Open in CodeRabbit

Reviews can continue after your included limit without a manual trigger. An admin must approve usage-based billing.

Or wait 14 minutes for your next included review.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available. Your 66 included PR review attempts over the past 7 days set your current allowance at 1 review per hour.

Learn how review limits work.

Review configuration:

Run configuration

Commits

Reviewing files that changed from the base of the PR and between b82691b and 88e24f4.

Files selected for processing: 4

  • packages/bash/build.ncl
  • packages/libtool/build.ncl
  • packages/libxml2/build.ncl
  • packages/libxslt/build.ncl
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

@norrietaylor norrietaylor left a comment
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Scope matches the title. The diff adds four OutputBin entries and one comment, in the four packages that the body names. Clobber none found. The diff only adds lines. The recent history (#1436, #1411) touches other outputs. I did not run git log per file because the harness gave no shell. Correctness The globs match the install paths that the body cites. The allow_missing_interpreter use follows the repo pattern (curl-config, pcre2-config, freetype-config, jemalloc.sh). libtool-next-version gets /bin/sh from the existing bash runtime dep. The main risk is in bashbug. Upstream's make rule writes $(CCFLAGS) into the script, and our CFLAGS holds -ffile-prefix-map=$(pwd)=/builddir. Thus the shipped script contains the absolute sandbox build path. The xml2-config/xslt-config consumer behaviour change is real, and the PR body states it. The world rebuild is the right check for it. Not checked (no build, no shell): the bytes of the built bashbug, xml2-config and xslt-config. I also did not check whether consumers such as lxml or PHP now pick up xml2-config before pkg-config.

Comment thread packages/bash/build.ncl
bash = { glob = "usr/bin/bash" } | OutputBin,
sh = { glob = "usr/bin/sh" } | OutputBin,
# #!/bin/sh script; the sh it needs is this package, which cannot be its own runtime dep.
bashbug = { glob = "usr/bin/bashbug", allow_missing_interpreter = true } | OutputBin,
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

low · bashbug now ships, and upstream bakes the compile flags into it. The bashbug make rule replaces !CFLAGS! in support/bashbug.sh with $(CCFLAGS). Our build.sh puts -ffile-prefix-map=$(pwd)=/builddir into CFLAGS. So the shipped script contains the absolute sandbox build directory, which is the path that flag exists to remove. If that directory is not the same for every build, bashbug is not byte-reproducible, and it sits in base. Check one built output: grep -n CFLAGS usr/bin/bashbug. If the path is there, rewrite it after install in build.sh, for example sed -i "s|$(pwd)|/builddir|g" "$OUTPUT_DIR/usr/bin/bashbug". No other package in the repo scrubs this. This is the only instance in the diff.

outputs = {
xmllint = { glob = "usr/bin/xmllint" } | OutputBin,
xmlcatalog = { glob = "usr/bin/xmlcatalog" } | OutputBin,
"xml2-config" = { glob = "usr/bin/xml2-config", allow_missing_interpreter = true } | OutputBin,
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit · xml2-config here and xslt-config in packages/libxslt/build.ncl:39 turn off the interpreter lint without a reason. The bashbug line in this PR has one. Most single-script siblings also have one, for example packages/nspr/build.ncl and packages/libgcrypt/build.ncl:41. Add one line, for example # #!/bin/sh script; consumers are builds, which always have a shell.

@bryan-minimal
bryan-minimal added this pull request to the merge queue Nov 13, 2026
Merged via the queue into main with commit ffa70ca Nov 13, 2026
15 checks passed
@bryan-minimal
bryan-minimal deleted the bin-coverage-low branch November 13, 2026 07:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants

0