#include "references.hh"
#include "hash.hh"
#include "util.hh"
#include "archive.hh"

#include <map>
#include <cstdlib>


namespace nix {


static unsigned int refLength = 32; /* characters */


static void search(const unsigned char * s, unsigned int len, 
    StringSet & hashes, StringSet & seen)
{
    static bool initialised = false;
    static bool isBase32[256];
    if (!initialised) {
        for (unsigned int i = 0; i < 256; ++i) isBase32[i] = false;
        for (unsigned int i = 0; i < base32Chars.size(); ++i)
            isBase32[(unsigned char) base32Chars[i]] = true;
        initialised = true;
    }
    
    for (unsigned int i = 0; i + refLength <= len; ) {
        int j;
        bool match = true;
        for (j = refLength - 1; j >= 0; --j)
            if (!isBase32[(unsigned char) s[i + j]]) {
                i += j + 1;
                match = false;
                break;
            }
        if (!match) continue;
        string ref((const char *) s + i, refLength);
        if (hashes.find(ref) != hashes.end()) {
            debug(format("found reference to `%1%' at offset `%2%'")
                  % ref % i);
            seen.insert(ref);
            hashes.erase(ref);
        }
        ++i;
    }
}


struct RefScanSink : Sink
{
    HashSink hashSink;
    StringSet hashes;
    StringSet seen;

    string tail;

    RefScanSink() : hashSink(htSHA256) { }
    
    void operator () (const unsigned char * data, size_t len);
};


void RefScanSink::operator () (const unsigned char * data, size_t len)
{
    hashSink(data, len);

    /* It's possible that a reference spans the previous and current
       fragment, so search in the concatenation of the tail of the
       previous fragment and the start of the current fragment. */
    string s = tail + string((const char *) data, len > refLength ? refLength : len);
    search((const unsigned char *) s.data(), s.size(), hashes, seen);

    search(data, len, hashes, seen);

    unsigned int tailLen = len <= refLength ? len : refLength;
    tail =
        string(tail, tail.size() < refLength - tailLen ? 0 : tail.size() - (refLength - tailLen)) +
        string((const char *) data + len - tailLen, tailLen);
}


PathSet scanForReferences(const string & path,
    const PathSet & refs, HashResult & hash)
{
    RefScanSink sink;
    std::map<string, Path> backMap;

    /* For efficiency (and a higher hit rate), just search for the
       hash part of the file name.  (This assumes that all references
       have the form `HASH-bla'). */
    foreach (PathSet::const_iterator, i, refs) {
        string baseName = baseNameOf(*i);
        string::size_type pos = baseName.find('-');
        if (pos == string::npos)
            throw Error(format("bad reference `%1%'") % *i);
        string s = string(baseName, 0, pos);
        assert(s.size() == refLength);
        assert(backMap.find(s) == backMap.end());
        // parseHash(htSHA256, s);
        sink.hashes.insert(s);
        backMap[s] = *i;
    }

    /* Look for the hashes in the NAR dump of the path. */
    dumpPath(path, sink);

    /* Map the hashes found back to their store paths. */
    PathSet found;
    foreach (StringSet::iterator, i, sink.seen) {
        std::map<string, Path>::iterator j;
        if ((j = backMap.find(*i)) == backMap.end()) abort();
        found.insert(j->second);
    }

    hash = sink.hashSink.finish();
        
    return found;
}


}
ltip'>Fixes &lt;https://bugs.gnu.org/45453&gt;.

* guix/import/utils.scm (alist-&gt;package): Find plain license names in (guix
licenses) first, and only fall back to SPDX names on error.
* tests/import-utils.scm ("alist-&gt;package with SPDX license name 1/2",
"alist-&gt;package with SPDX license name 2/2"): New tests.
</span></span></td><td>Ricardo Wurmus</td></tr>
<tr><td><span title='2020-12-02 22:09:23 +0100'>2020-12-02</span></td><td><a href='/guix/commit/tests/import-utils.scm?id=bea3b17739fc591b8cf6db1f8d28a6f6c9585577'>import: utils: 'recursive-import' accepts an optional version parameter.</a><span class='msg-avail'>...<span class='msg-tooltip'>This adds a key VERSION to 'recursive-import' and moves the parameter REPO to
a key. This also changes all the places that rely on 'recursive-import'.

* guix/import/utils.scm (recursive-import): Add the VERSION key. Make REPO a
  key.
  (package-&gt;definition): Add optional 'append-version?'.
* guix/scripts/import/crate.scm (guix-import-crate): Add the VERSION key.
* guix/import/crate.scm (crate-&gt;guix-package): Add the VERSION key.
  (crate-recursive-import): Pass VERSION to recursive-import, remove now
  unnecessary code.
* guix/import/cran.scm (cran-&gt;guix-package, cran-recursive-import): Change the
  REPO parameter to a key.
* guix/import/elpa.scm (elpa-&gt;guix-package, elpa-recursive-import): Likewise.
* guix/import/gem.scm (gem-&gt;guix-package, recursive-import): Likewise.
* guix/import/opam.scm (opam-recurive-import): Likewise.
* guix/import/pypi.scm (pypi-recursive-import): Likewise.
* guix/import/stackage.scm (stackage-recursive-import): Likewise.
* guix/scripts/import/cran.scm (guix-import-cran): Likewise.
* guix/scripts/import/elpa.scm (guix-import-elpa): Likewise.
* tests/elpa.scm (eval-test-with-elpa): Likewise.
* tests/import-utils.scm (recursive-import): Likewise.

Co-authored-by: Hartmut Goebel &lt;h.goebel@crazy-compilers.com&gt;
</span></span></td><td>Martin Becze</td></tr>