0003767: Zip methods process files in an unpredictable order, and don't process all files

(0004296)
dobes_vandermeer
11-16-09 16:57

Note that this is using the version of Quercus included in Resin 4.0.1. I wasn't able to select 4.0.1 as the version, unfortunately.

(0004298)
dobes_vandermeer
11-16-09 20:56

What did work is to access a java class that extracts the files for me.

Here's the class:

import java.io.File;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InputStream;
import java.util.Enumeration;
import java.util.zip.ZipEntry;
import java.util.zip.ZipFile;

import org.apache.commons.io.IOUtils;

/**
* Extract a ZIP file into some folder.
*/
public class ZipExtractor {
    public static void extract(String zip, String outputFolder) throws IOException {
        ZipFile zf = new ZipFile(zip);
        File of = new File(outputFolder).getAbsoluteFile();
        Enumeration<? extends ZipEntry> ee = zf.entries();
        while(ee.hasMoreElements()) {
            ZipEntry ze = ee.nextElement();
            String name = ze.getName();
            File outfile = new File(of, name);
            if(ze.isDirectory()) {
                outfile.mkdirs();
            } else {
                InputStream in = zf.getInputStream(ze);
                outfile.getParentFile().mkdirs(); // Make sure the folder exists
                FileOutputStream out = new FileOutputStream(outfile);
                IOUtils.copy(in, out);
            }
        }
    }
}

Here's the way it can be invoked from PHP code in Quercus:

function unzip_file($file, $to) {
    $util = java_class('books.util.ZipExtractor');
    $util->extract($file, $to);
    return true;
}

So ... if anyone else is looking for a workaround, here you go!

(0005065)
doogie
02-17-11 10:45

At a guess, when looking at that zip file, the zone_grad.gif is the first entry in the file that is actually *expanded* when it was added to the zip. It's compressed size is correctly reported as 114, but the real uncompressed size is supposed to be 63, *not* 148!.

At a guess, this might be due to an abs() in the code(unlikely), or a signed vs. unsigned int problem(more likely). I haven't actually tried to dig into this codebase at all, just looking at the debug above and the zip file(unzip -lv).

Also, once quercus thinks the file size is 148(which is too much), it then starts to overflow a buffer or something.

(0005066)
doogie
02-17-11 11:09

ZipModule.zip_entry_filesize returns Value, and uses LongValue.create() internally. ZipMmodule.zip_entry_compressedsize returns long directly. Maybe that is the problem.

I really think those methods should be consistent.

(0005916)
nam
06-21-12 01:14

php/1v0z

Fixed for 4.0.29.

Issue History
Date Modified	Username	Field	Change
11-16-09 16:55	dobes_vandermeer	New Issue
11-16-09 16:55	dobes_vandermeer	File Added: fresh-page.1.1.zip
11-16-09 16:57	dobes_vandermeer	Note Added: 0004296
11-16-09 20:56	dobes_vandermeer	Note Added: 0004298
02-17-11 10:45	doogie	Note Added: 0005065
02-17-11 11:09	doogie	Note Added: 0005066
02-18-11 15:00	doogie	Issue Monitored: doogie
02-26-11 15:04	kdecherf	Issue Monitored: kdecherf
06-21-12 01:13	nam	Status	new => assigned
06-21-12 01:13	nam	Assigned To	=> nam
06-21-12 01:14	nam	Status	assigned => closed
06-21-12 01:14	nam	Note Added: 0005916
06-21-12 01:14	nam	Resolution	open => fixed

Notes
(0004296) dobes_vandermeer 11-16-09 16:57	Note that this is using the version of Quercus included in Resin 4.0.1. I wasn't able to select 4.0.1 as the version, unfortunately.

(0004298) dobes_vandermeer 11-16-09 20:56	What did work is to access a java class that extracts the files for me. Here's the class: import java.io.File; import java.io.FileOutputStream; import java.io.IOException; import java.io.InputStream; import java.util.Enumeration; import java.util.zip.ZipEntry; import java.util.zip.ZipFile; import org.apache.commons.io.IOUtils; /** * Extract a ZIP file into some folder. */ public class ZipExtractor { public static void extract(String zip, String outputFolder) throws IOException { ZipFile zf = new ZipFile(zip); File of = new File(outputFolder).getAbsoluteFile(); Enumeration<? extends ZipEntry> ee = zf.entries(); while(ee.hasMoreElements()) { ZipEntry ze = ee.nextElement(); String name = ze.getName(); File outfile = new File(of, name); if(ze.isDirectory()) { outfile.mkdirs(); } else { InputStream in = zf.getInputStream(ze); outfile.getParentFile().mkdirs(); // Make sure the folder exists FileOutputStream out = new FileOutputStream(outfile); IOUtils.copy(in, out); } } } } Here's the way it can be invoked from PHP code in Quercus: function unzip_file($file, $to) { $util = java_class('books.util.ZipExtractor'); $util->extract($file, $to); return true; } So ... if anyone else is looking for a workaround, here you go!

(0005065) doogie 02-17-11 10:45	At a guess, when looking at that zip file, the zone_grad.gif is the first entry in the file that is actually expanded when it was added to the zip. It's compressed size is correctly reported as 114, but the real uncompressed size is supposed to be 63, not 148!. At a guess, this might be due to an abs() in the code(unlikely), or a signed vs. unsigned int problem(more likely). I haven't actually tried to dig into this codebase at all, just looking at the debug above and the zip file(unzip -lv). Also, once quercus thinks the file size is 148(which is too much), it then starts to overflow a buffer or something.

(0005066) doogie 02-17-11 11:09	ZipModule.zip_entry_filesize returns Value, and uses LongValue.create() internally. ZipMmodule.zip_entry_compressedsize returns long directly. Maybe that is the problem. I really think those methods should be consistent.

(0005916) nam 06-21-12 01:14	php/1v0z Fixed for 4.0.29.

Relationships