bug#22333: 24.5; EWW downloads invalid compressed tar-files

bug#22333: 24.5; EWW downloads invalid compressed tar-files

[Iñigo Serna]

Hello,

Browsing web using eww and downloading compressed tar-files
(f.e. filename.tar.gz) save invalid files (can't be uncompressed).

I've tested on emacs 24.5 official packages both in Fedora 23 (x86_64)
and ArchLinux (x86_64 and arm), running with "emacs -Q -nw".
Tried also on git master, but download fails with this message:

    error in process filter: Args out of range: #<buffer  *http pypi.python.org:443*>, 0, 13

To reproduce, browse to any web which offers any tar compressed file
and hit 'd' to download when cursor is over file URL. Then try to
uncompress it:

    /home/inigo/Downloads ⚡ tar xvf filename.tar.gz
    tar: This does not look like a tar archive
    tar: Skipping to next header
    tar: Exiting with failure status due to previous errors

I see `eww-download` calls `url-retrieve` to download the file
asynchronously, which stores the downloaded file in an emacs buffer and
then saves it.
I think the problem comes from the buffer saving action and not from eww
code, but I don't have enough skills to debug this.

Btw, using `url-copy-file` (which does not employ an intermediate emacs
buffer) works ok, so I use this code:

    (defun isr/eww-download ()
      "Alternative not asynchronous download."
      (interactive)
       (let ((url (get-text-property (point) 'shr-url)))
         (unless url
           (message "No URL under point"))
         (url-copy-file url (concat eww-download-directory "/" (file-name-nondirectory url)))
         (message "Saved %s" url)))

Thanks,
Iñigo Serna

PS: This is my first emacs bug report, excuse me if I've done something wrongly.

INFORMATION COPIED FROM report-emacs-bug:

Configured using:
 `configure --build=x86_64-redhat-linux-gnu
 --host=x86_64-redhat-linux-gnu --program-prefix=
 --disable-dependency-tracking --prefix=/usr --exec-prefix=/usr
 --bindir=/usr/bin --sbindir=/usr/sbin --sysconfdir=/etc
 --datadir=/usr/share --includedir=/usr/include --libdir=/usr/lib64
 --libexecdir=/usr/libexec --localstatedir=/var
 --sharedstatedir=/var/lib --mandir=/usr/share/man
 --infodir=/usr/share/info --with-dbus --with-gif --with-jpeg --with-png
 --with-rsvg --with-tiff --with-xft --with-xpm --with-x-toolkit=gtk3
 --with-gpm=no build_alias=x86_64-redhat-linux-gnu
 host_alias=x86_64-redhat-linux-gnu 'CFLAGS=-DMAIL_USE_LOCKF -O2 -g
 -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2
 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4
 -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1
 -m64 -mtune=generic' LDFLAGS=-Wl,-z,relro'

Important settings:
  value of $LANG: en_GB.UTF-8
  locale-coding-system: utf-8-unix

bug#22333: 24.5; EWW downloads invalid compressed tar-files

[Eli Zaretskii]

> From: Iñigo Serna <inigoserna@gmail.com>
> Date: Fri, 08 Jan 2016 22:23:27 +0100
> 
> Browsing web using eww and downloading compressed tar-files
> (f.e. filename.tar.gz) save invalid files (can't be uncompressed).
> 
> I've tested on emacs 24.5 official packages both in Fedora 23 (x86_64)
> and ArchLinux (x86_64 and arm), running with "emacs -Q -nw".
> Tried also on git master, but download fails with this message:
> 
>     error in process filter: Args out of range: #<buffer  *http pypi.python.org:443*>, 0, 13
> 
> To reproduce, browse to any web which offers any tar compressed file
> and hit 'd' to download when cursor is over file URL. Then try to
> uncompress it:
> 
>     /home/inigo/Downloads ⚡ tar xvf filename.tar.gz
>     tar: This does not look like a tar archive
>     tar: Skipping to next header
>     tar: Exiting with failure status due to previous errors

Could it be that the saved tar file is already uncompressed?

bug#22333: 24.5; EWW downloads invalid compressed tar-files

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

>> To reproduce, browse to any web which offers any tar compressed file
>> and hit 'd' to download when cursor is over file URL. Then try to
>> uncompress it:
>> 
>>     /home/inigo/Downloads ⚡ tar xvf filename.tar.gz
>>     tar: This does not look like a tar archive
>>     tar: Skipping to next header
>>     tar: Exiting with failure status due to previous errors
>
> Could it be that the saved tar file is already uncompressed?

I'd rather guess it's compressed twice.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#22333: 24.5; EWW downloads invalid compressed tar-files

[Iñigo Serna]

Hi and thanks for your time,

Andreas Schwab writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>> Could it be that the saved tar file is already uncompressed?
>
> I'd rather guess it's compressed twice.

I think I've discovered the problem with the downloaded files:

0. As downloaded (remember it's saved from an emacs buffer) is not a
valid .tar.gz file
1. but it is a valid gzipped file
2. the uncompressed file contains some HTTP headers at
the start of the file.
3. After manually removing these lines and changing extension (to add
the '.gz' again), file is correctly recognized as a gzipped tar-file.

[0] /home/inigo/Downloads ⚡ tar xvfz filename.tar.gz 
    tar: This does not look like a tar archive
    tar: Skipping to next header
    tar: Exiting with failure status due to previous errors

[1] /home/inigo/Downloads ⚡ gzip -d filename.tar.gz 
    /home/inigo/Downloads ⚡ file filename.tar 
    filename.tar: data

[2] /home/inigo/Downloads ⚡ head -21 filename.tar 
    HTTP/1.1 200 OK
    x-amz-replication-status: COMPLETED
    Last-Modified: Sun, 08 Nov 2015 13:11:05 GMT
    ETag: "9c13c5fafcb1aecd43f51fa9b0278000"
    Content-Type: application/octet-stream
    Server: AmazonS3
    Via: 1.1 varnish
    Fastly-Debug-Digest: 37a3779f444d206796099a174bb873c853bb8f1b9a13cf06f29108f907a9d50b
    Cache-Control: max-age=31557600, public
    Content-Length: 87341
    Accept-Ranges: bytes
    Date: Sat, 09 Jan 2016 10:16:29 GMT
    Via: 1.1 varnish
    Age: 135058
    Connection: keep-alive
    X-Served-By: cache-sea1920-SEA, cache-fra1247-FRA
    X-Cache: HIT, HIT
    X-Cache-Hits: 2, 1
    X-Timer: S1452334589.874408,VS0,VE1
    
    *I?V.�dist/lfm-3.0.tar��[w۸�0���.�d���H�$;�>:����N�+��c;}9n/

[3] [...remove HTTP headers from first lines of file...]
    /home/inigo/Downloads ⚡ file filename.tar 
    filename.tar: gzip compressed data, was "dist/lfm-3.0.tar", last modified: Sun Nov  8 14:07:54 2015, max compression
    /home/inigo/Downloads ⚡ mv filename.tar filename.tar.gz
    /home/inigo/Downloads ⚡ file filename.tar.gz 
    filename.tar.gz: gzip compressed data, was "dist/lfm-3.0.tar", last modified: Sun Nov  8 14:07:54 2015, max compression
    /home/inigo/Downloads ⚡ tar xvfz filename.tar.gz 
    lfm-3.0/
    [...]
    

So I think the error comes with those extra spurious headers added to
the start of the buffer when downloading the file, which mess buffer
saving operation.

The bug can be easily reproduced with code like:

    (let ((url "https://pypi.python.org/packages/source/l/lfm/lfm-3.0.1.tar.gz"))
      (url-retrieve url 'test-cb (list url)))
    
    (defun test-cb (status url)
      (message "Downloaded: %s" url)
      ;; (sleep-for 60)
      (write-file "~/a.tgz")
      (message "Saved: %s" url))
    

`eww-download-callback` (called by `url-retrieve`) should remove those headers 
after download is finished and before buffer is saved to disk.

A simple fix could be to add (excuse my poor elisp skills):

    (goto-char (point-min))
    (search-forward-regexp "^$")
    (forward-line)
    (delete-region (point-min) (point))

before '(write-file file)' in `eww-download-callback`.

... and this almost works, but the downloaded file is compressed twice
now as Andreas stated (ie, tar.gz is gzipped again at file saving).

Thus it looks there is an additional issue when saving gzipped tar-files
from a buffer.

Any idea on how to solve this?


Thanks,
Iñigo Serna

bug#22333: 24.5; EWW downloads invalid compressed tar-files

[Lars Ingebrigtsen]

Iñigo Serna <inigoserna@gmail.com> writes:

> `eww-download-callback` (called by `url-retrieve`) should remove those headers 
> after download is finished and before buffer is saved to disk.
>
> A simple fix could be to add (excuse my poor elisp skills):
>
>     (goto-char (point-min))
>     (search-forward-regexp "^$")
>     (forward-line)
>     (delete-region (point-min) (point))
>
> before '(write-file file)' in `eww-download-callback`.

This function looks like this in emacs-25:

(defun eww-download-callback (status url)
  (unless (plist-get status :error)
    (let* ((obj (url-generic-parse-url url))
           (path (car (url-path-and-query obj)))
           (file (eww-make-unique-file-name
                  (eww-decode-url-file-name (file-name-nondirectory path))
                  eww-download-directory)))
      (goto-char (point-min))
      (re-search-forward "\r?\n\r?\n")
      (write-region (point) (point-max) file)
      (message "Saved %s" file))))

So it looks like this has already been fixed.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#22333: 24.5; EWW downloads invalid compressed tar-files

[Lars Ingebrigtsen]

Iñigo Serna <inigoserna@gmail.com> writes:

> ... and this almost works, but the downloaded file is compressed twice
> now as Andreas stated (ie, tar.gz is gzipped again at file saving).

And I've confirmed that the files are not double-gzipped when hitting
`d' on links on the following pages:

http://www-ftp.lip6.fr/pub/emacs/gnus/

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#22333: 24.5; EWW downloads invalid compressed tar-files

[Iñigo Serna]

Hi

Lars Ingebrigtsen writes:
> And I've confirmed that the files are not double-gzipped when hitting
> `d' on links on the following pages:
>
> http://www-ftp.lip6.fr/pub/emacs/gnus/

I confirm it works perfectly with every link I've tried.

Thanks,
Iñigo Serna