Repository navigation
Round10 fixes - #37
Merged
Merged
Conversation
- Open SQLite databases at the path the operating system resolves, as filepath.Abs cleans .. against the shell's working directory, which wrote an empty output and could audit against a different database when the working directory was reached through a symlink - Index SQLite records containing .. or .// rather than looking them up directly, which never matched them - Keep spaces at the ends of the hashdeep invoked from directory - Only remove a carriage return from hashdeep lines with windows line endings, so filenames ending in one can be audited - Add round 10 checks to test-all.sh Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- A path containing .. has everything up to its last .. resolved by the operating system when comparing paths, so after a symlink it names the directory actually read. Auditing ../d from a symlinked working directory, or a supplied path such as link/../d, could compare the files read against the records of a different directory and pass a modified tree - Supplied paths are only cleaned when that names the same directory, and --absolute and -i resolve .. the same way - A sum file listing a file more than once has every line checked as md5sum -c does, a different value of the same hash reporting it modified - List an unreadable file whose SQLite record was paired as a move - README: find -L, symlinked working directories, sum duplicates, SQLite index - Add round 11 checks to test-all.sh Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ates through every directory Resolving every .. through the operating system mixed physical and logical paths from a symlinked working directory, reporting every file as moved, and a .. the operating system cannot follow was cleaned into another path, so could match a different file. Only the path up to the last .. of a supplied path is resolved, so a symlink after it is walked as supplied. Sum file lines naming the same file only through the sum file's directory are now merged, so are all checked. SQLite audits report every spelling of an unreadable file and list unreadable files in order. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…uments and records A manifest whose invoked from directory was deleted or made unreadable reported every file as moved, as .. through it was kept as written. A directory which no longer exists cannot be a symlink, so .. after it is removed as text, while .. after a dangling link or a file still never matches. Paths supplied with .. now use the same rule as paths in audit files, so they match hashdeep manifests made through symlinked directories. Sum file lines are compared across every directory by the file they name, so conflicting lines are found through symlinked and different directories, and directories are not files. A missing SQLite record is only reported unreadable through its own path. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…h unsearchable directories Merging sum lines naming the same file dropped a line's own path, so a scanned symlink to a listed file was new, and let the kept line be checked against a different file than the dropped line named. Every line is now kept, lines naming the same real file are grouped for conflicts, and a line is only not reported when another was checked against that same file. A directory which cannot be searched may hold a symlink, so .. after it never matches rather than being cleaned as text. A physical path inside the working directory or a directory containing it is given through the working directory, so lnk/../.. matches files found from a symlinked working directory. SQLite -v finds unreadable spellings through the first directory which exists. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…, check every hash of a sum file's lines for one file Rewriting only the part of a path up to its last .. through the symlinked working directory gave one file two keys, depending on whether it was reached by an absolute physical path or by .. landing above it. Every compared path inside where the working directory, or a directory containing it, really is, is now compared through it, which also lets an absolute physical manifest pass when audited from the symlink. SQLite lookups also try the physical spelling, with as many placeholders as there are candidates. Sum file lines naming the same file are all checked with every hash listed for it, so a line with another algorithm is not hidden as a duplicate. SQLite -v checks other spellings of an unreadable file through the directory its record was found through. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…spelling, keep conflicts through merges Rewriting paths into a symlinked working directory was not canonical: with nested symlinked ancestors a path written through one could get another key, an unclean $PWD such as // applied it twice inside symlink loops, and cleaning the joined path matched x/.. paths which can never be opened. Paths are now taken through where every symlinked ancestor really is until nothing changes, then through the deepest containing it, joined as text. When the working directory was reached through a symlink every SQLite record is indexed, as a record can be compared as another spelling. A sum line's conflicting value is kept when its record is merged or grouped with another spelling of the file, so the line is not hidden as a duplicate. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…lled SQLite records A working directory reached through a symlink to / mapped /tmp/x to W/Rtmp/x, which could match a SQLite record of that literal path. When the working directory was reached through a symlink only records whose compared path differs from their spelling are indexed, rather than every record. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An audit file given as x/lnk/../m.hd added x as a directory relative records are resolved against, as filepath.Dir cleaned the path as text, so a record could match a different file in x. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ame path with the same values Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…r directory places its record is reported as moved Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ough, chosen by most files Sum lines only stand in for a checked line when they name the checked file through that directory, records dropped as duplicates are audited when they name another file there, and SQLite duplicates are kept by row. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…fuse SQLite databases not keyed by path A sum file listing every file by relative and absolute path, audited in place from another directory, matched only absolute lines so no directory was chosen, and its relative lines were judged through the working directory and reported as missing. A SQLite table without filepath as its primary key could list a path twice with only one audited, and a rowid column hid the real rowid. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…was chosen, and refuse generated rowid columns and inexact path keys Grouping rows through every directory let a relative row naming a deleted file through the directory hashit was invoked from be dropped as a duplicate of an absolute row naming another file through the working directory. Only sum files, which record no such directory, need that. The schema check missed a generated rowid column, which pragma_table_info does not list, and a filepath key with a collation such as nocase, which let a renamed file pass. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ctly, or is missing on a row A collation on the filepath column, with the key itself compared exactly, let a rename differing in case or trailing space pass, and a numeric filepath type let a file renamed from 1 to 01 or 1.0 pass, as lookups compare using the column. A row without a path failed with a raw scan error. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
pr-insights
Bot
added
M/size
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
So many fixes..