Debian Bug report logs - #12240
dpkg-deb --build doesn't obey $TMPDIR

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: "Mark W. Eichin" <eichin@kitten.gen.ma.us>

To: submit@bugs.debian.org

Subject: dpkg-deb --build doesn't obey $TMPDIR

Date: Sat, 23 Aug 1997 15:00:48 -0400

Package: dpkg
Version: 1.4.0.19

% ( setenv TMPDIR /var/bigtmp ; dpkg --build debian/tmp-xbooks .. )
dpkg-deb: building package `xbooks' in `../xbooks_3.3-5_all.deb'.

gzip: stdout: No space left on device
dpkg-deb: subprocess gzip -9c from tar --exclude returned error exit status 1

strace shows it using /tmp; /tmp has about 2M, /var/bigtmp has about 150M...

Message #10 received at 12240@bugs.debian.org (full text, mbox, reply):

From: Juan Cespedes <cespedes@debian.org>

To: "Mark W. Eichin" <eichin@kitten.gen.ma.us>, 12240@bugs.debian.org

Cc: control@bugs.debian.org, libc6@packages.debian.org

Subject: Re: dpkg-deb --build doesn't obey $TMPDIR

Date: Sun, 5 Apr 1998 14:06:08 +0200

reassign 12240 libc6
retitle 12240 tmpdir,tmpnam should obey $TMPDIR
thanks

From: "Mark W. Eichin" <eichin@kitten.gen.ma.us>
> % ( setenv TMPDIR /var/bigtmp ; dpkg --build debian/tmp-xbooks .. )
> dpkg-deb: building package `xbooks' in `../xbooks_3.3-5_all.deb'.
> 
> gzip: stdout: No space left on device
> dpkg-deb: subprocess gzip -9c from tar --exclude returned error exit status 1
> 
> strace shows it using /tmp; /tmp has about 2M, /var/bigtmp has about 150M...

	`dpkg-deb' uses tmpdir() and tmpnam() to get access to temp
files, so I'm reassigning this bug to libc6.

	Thanks.


-- 
Juan Cespedes

Message #19 received at 12240@bugs.debian.org (full text, mbox, reply):

From: James Troup <J.J.Troup@scm.brad.ac.uk>

To: Juan Cespedes <cespedes@debian.org>

Cc: 12240@bugs.debian.org, "Mark W. Eichin" <eichin@kitten.gen.ma.us>, libc6@packages.debian.org

Subject: Re: Bug#12240: dpkg-deb --build doesn't obey $TMPDIR

Date: 06 Apr 1998 15:39:32 +0100

reassign 12240 dpkg
retitle 12240 dpkg-deb --build doesn't obey $TMPDIR
thanks

(Reassigned back because you're not the maintainer and the reason you
gave for reassignment was bgous.  I'm not the maintainer either (or
even anything to with dpkg), but I'm just restoring things to how they
were)

Juan Cespedes <cespedes@debian.org> writes:

> reassign 12240 libc6
> retitle 12240 tmpdir,tmpnam should obey $TMPDIR
> thanks
> 
> From: "Mark W. Eichin" <eichin@kitten.gen.ma.us>
> > % ( setenv TMPDIR /var/bigtmp ; dpkg --build debian/tmp-xbooks .. )
> > dpkg-deb: building package `xbooks' in `../xbooks_3.3-5_all.deb'.
> > 
> > gzip: stdout: No space left on device
> > dpkg-deb: subprocess gzip -9c from tar --exclude returned error exit status 1
> > 
> > strace shows it using /tmp; /tmp has about 2M, /var/bigtmp has about 150M...
> 
> 	`dpkg-deb' uses tmpdir() and tmpnam() to get access to temp
> files, so I'm reassigning this bug to libc6.

No it does not[1].  When building dpkg-deb uses tmpfile(), which uses
a hardcoded path (in <stdio.h>)[2] for finding the tmp directory and
this is the documented behaviour of that function.  Presumably Mark
wants dpkg-deb to use something which obeys $TMPDIR, e.g. tempnam.

If you think the API of tmpfile and friends should change, please open
a separate bug report against libc6.

Juan, WTF are you doing?

[1] 
15:04:08@dcsun4| ~/temp/dpkg-1.4.0.21/dpkg-deb $grep tmp *.c
build.c:  if (!(gz= tmpfile())) ohshite(_("failed to make tmpfile (control)"));
build.c:  if (fstat(fileno(gz),&controlstab)) ohshite(_("failed to fstat tmpfile (control)"));
build.c:  if (lseek(fileno(gz),0,SEEK_SET)) ohshite(_("failed to rewind tmpfile (control)"));
build.c:    if (!(gz= tmpfile())) ohshite(_("failed to make tmpfile (data)"));
build.c:    if (fstat(fileno(gz),&datastab)) ohshite("_(failed to fstat tmpfile (data))");
build.c:    if (lseek(fileno(gz),0,SEEK_SET)) ohshite(_("failed to rewind tmpfile (data)"));
info.c:  static char dbuf[L_tmpnam];
info.c:  if (!tmpnam(dbuf)) ohshite(_("failed to make temporary filename"));
15:04:11@dcsun4| ~/temp/dpkg-1.4.0.21/dpkg-deb $

[2] 
| DESCRIPTION
|       The  tmpfile() function generates a unique temporary file­
|       name using the path prefix P_tmpdir defined in  <stdio.h>.
                                                        -- tmpfile(3)

-- 
James

Message #28 received at 12240@bugs.debian.org (full text, mbox, reply):

From: Topi Miettinen <Topi.Miettinen@medialab.sonera.net>

To: 12240@bugs.debian.org

Subject: patch

Date: Fri, 03 Jul 1998 23:04:49 +0300

This patch attempts to fix the problem.

diff -ru dpkg-1.4.0.23.2/dpkg-deb/build.c.orig dpkg-1.4.0.23.2/dpkg-deb/build.c
--- dpkg-1.4.0.23.2/dpkg-deb/build.c.orig       Sat Mar  8 21:40:47 1997
+++ dpkg-1.4.0.23.2/dpkg-deb/build.c    Fri Jul  3 22:45:48 1998
@@ -57,13 +57,13 @@
     PREINSTFILE, POSTINSTFILE, PRERMFILE, POSTRMFILE, 0
   };
   
-  char *m;
+  char *m, *tmpd, *tmpf;
   const char *debar, *directory, *const *mscriptp, *versionstring, *arch;
   char *controlfile;
   struct pkginfo *checkedinfo;
   struct arbitraryfield *field;
   FILE *ar, *gz, *cf;
-  int p1[2],p2[2], warns, errs, n, c, subdir;
+  int p1[2],p2[2], warns, errs, n, c, subdir, gzfd;
   pid_t c1,c2,c3,c4,c5;
   struct stat controlstab, datastab, mscriptstab, debarstab;
   char conffilename[MAXCONFFILENAME+1];
@@ -208,7 +208,14 @@
     execlp(TAR,"tar","-cf","-",".",(char*)0); ohshite(_("failed to exec tar -cf"));
   }
   close(p1[1]);
-  if (!(gz= tmpfile())) ohshite(_("failed to make tmpfile (control)"));
+
+  if (!(tmpd = getenv("TMPDIR")))
+    tmpd= "/tmp";
+  tmpf= malloc(strlen(tmpd) + strlen("/dpkg.XXXXXX"));
+  strcpy(tmpf, tmpd);
+  strcat(tmpf, "/dpkg.XXXXXX");
+  if (!(gzfd= mkstemp(tmpf)) || !(gz= fdopen(gzfd, "r+")))
+    ohshite(_("failed to make tmpfile (control)"));
   if (!(c2= m_fork())) {
     m_dup2(p1[0],0); m_dup2(fileno(gz),1); close(p1[0]);
     execlp(GZIP,"gzip","-9c",(char*)0); ohshite(_("failed to exec gzip -9c"));
@@ -245,8 +252,12 @@
   
   if (!oldformatflag) {
     fclose(gz);
-    if (!(gz= tmpfile())) ohshite(_("failed to make tmpfile (data)"));
+    strcpy(tmpf, tmpd);
+    strcat(tmpf, "/dpkg.XXXXXX");
+    if (!(gzfd= mkstemp(tmpf)) || !(gz= fdopen(gzfd, "r+")))
+      ohshite(_("failed to make tmpfile (data)"));
   }
+  free(tmpf);
   m_pipe(p2);
   if (!(c4= m_fork())) {
     m_dup2(p2[1],1); close(p2[0]); close(p2[1]);
diff -ru dpkg-1.4.0.23.2/dpkg-deb/info.c.orig dpkg-1.4.0.23.2/dpkg-deb/info.c
--- dpkg-1.4.0.23.2/dpkg-deb/info.c.orig        Sat Mar  8 22:02:47 1997
+++ dpkg-1.4.0.23.2/dpkg-deb/info.c     Fri Jul  3 22:41:48 1998
@@ -59,12 +59,13 @@
                          const char **debarp,
                          const char **directoryp,
                          int admininfo) {
-  static char dbuf[L_tmpnam];
+  char *dbuf;
   pid_t c1;
   
   *debarp= *(*argvp)++;
   if (!*debarp) badusage(_("--%s needs a .deb filename argument"),cipaction->olong);
-  if (!tmpnam(dbuf)) ohshite(_("failed to make temporary filename"));
+  dbuf = tempnam(NULL, "dpkg");
+  if (!dbuf) ohshite(_("failed to make temporary filename"));
   *directoryp= dbuf;
 
   if (!(c1= m_fork())) {
@@ -73,6 +74,7 @@
   waitsubproc(c1,"rm -rf",0);
   push_cleanup(cu_info_prepare,-1, 0,0, 1, (void*)dbuf);
   extracthalf(*debarp, dbuf, "mx", admininfo);
+  free(dbuf);
 }
 
 static int ilist_select(const struct dirent *de) {

Message #33 received at 12240@bugs.debian.org (full text, mbox, reply):

From: Ian Jackson <ian@chiark.greenend.org.uk>

To: 12240@bugs.debian.org

Cc: Topi Miettinen <Topi.Miettinen@medialab.sonera.net>

Subject: Re: Bug#12240: patch

Date: Sat, 4 Jul 1998 14:24:12 +0100

The patch from Topi Miettinen should not be applied.  I agree with
Juan that this is a bug in the libc (though I also agree with James
Troup that he should probably not have reassigned the bug).

tmpfile() is documented in the glibc info file as:

 - Function: FILE * tmpfile (void)
     This function creates a temporary binary file for update mode, as
     if by calling `fopen' with mode `"wb+"'.  The file is deleted
     automatically when it is closed or when the program terminates.
     (On some other ANSI C systems the file may fail to be deleted if
     the program terminates abnormally).

Nowhere in this document does it say that tmpfile has the behaviour
that it's hardwired to use /tmp.  In fact, I think that such behaviour
is broken, and should be fixed.  I'm therefore going to reassign this
bug back to libc6.

I strongly suspect that the manpage for tmpfile was written with
reference to the implementation, rather than as a specification.  As
such it should not be considered canonical.  Even if it is to be
considered canonical, the behaviour documented in tmpfile(3) is still
broken, and so the documentation should be changed.  I'm therefore
also going to include manpages in the list of packages to which I'll
reassign the bug.

I'd be grateful if the manpages and/or libc maintainers would reassign
the bug to the other package when they have fixed their part of the
bug.

Thanks,
Ian.

Message #40 received at 12240@bugs.debian.org (full text, mbox, reply):

From: kaih@khms.westfalen.de (Kai Henningsen)

To: 12240@bugs.debian.org

Subject: Re: Bug#12240: patch

Date: 05 Jul 1998 11:37:00 +0200

ian@chiark.greenend.org.uk (Ian Jackson)  wrote on 04.07.98 in <E0ysSIW-0006Qg-00@chiark.greenend.org.uk>:

[...]
> tmpfile() is documented in the glibc info file as:
[...]

Just for reference, the following is the definition for tempfile() from  
Unix98:


                 The Single UNIX « Specification, Version 2
                      Copyright + 1997 The Open Group
     _________________________________________________________________

    NAME

     tmpfile - create a temporary file

    SYNOPSIS

#include <stdio.h>

FILE *tmpfile(void);

    DESCRIPTION

     The tmpfile() function creates a temporary file and opens a
     corresponding stream. The file will automatically be deleted when
     all references to the file are closed. The file is opened as in
     fopen() for update (w+).

     The largest value that can be represented correctly in an object of
     type off_t will be established as the offset maximum in the open
     file description.

     If the process is killed in the period between file creation and
     unlinking, a permanent file may be left behind.

     An error message may be written to standard error if the stream
     cannot be opened.

    RETURN VALUE

     Upon successful completion, tmpfile() returns a pointer to the
     stream of the file that is created. Otherwise, it returns a null
     pointer and sets errno to indicate the error.

    ERRORS

     The tmpfile() function will fail if:
   [EINTR]
          A signal was caught during tmpfile().
   [EMFILE]
          {OPEN_MAX} file descriptors are currently open in the calling
          process.
   [ENFILE]
          The maximum allowable number of files is currently open in the
          system.
   [ENOSPC]
          The directory or file system which would contain the new file
          cannot be expanded.
   [EOVERFLOW]
          The file is a regular file and the size of the file cannot be
          represented correctly in an object of type off_t.

     The tmpfile() function may fail if:
   [EMFILE]
          {FOPEN_MAX} streams are currently open in the calling process.
   [ENOMEM]
          Insufficient storage space is available.

    EXAMPLES

     None.

    APPLICATION USAGE

     None.

    FUTURE DIRECTIONS

     None.

    SEE ALSO

     fopen(), tmpnam(), unlink(), <stdio.h>.
     _________________________________________________________________

            UNIX « is a registered Trademark of The Open Group.
                      Copyright + 1997 The Open Group
              [ Main Index | XSH | XCU | XBD | XCURSES | XNS ]
     _________________________________________________________________

MfG Kai

Message #45 received at 12240@bugs.debian.org (full text, mbox, reply):

From: Topi Miettinen <Topi.Miettinen@medialab.sonera.net>

To: 12240@bugs.debian.org

Cc: Ian Jackson <ian@chiark.greenend.org.uk>

Subject: Re: Bug#12240: patch

Date: Tue, 07 Jul 1998 23:20:08 +0300

There are two parts in the patch, one for tmpfile() use and other for
tmpnam().

Regarding tmpfile(), I checked K&R, Stevens, Donald Lewine's "POSIX
Programmer's Guide", and Solaris manual page and they all agree with glibc
documentation in that nothing is guaranteed. Fixing this in glibc is
trivial, by changing fifth argument of __stdio_gen_tempname call from 0 to
1 in stdio-common/tmpfile.c. This will cause security problems with setuid
programs using tmpfile(), though. If glibc is fixed, the first part of the
patch affecting build.c could be ignored, but I fail to see other than
aesthetical problems with the patch.

However, tmpnam() is different, it is specified to use only L_tmpnam bytes
of user-specified buffer. Fixing glibc would therefore be non-trivial,
maybe like only using TMPDIR if the whole string would fit in L_tmpnam
bytes. Currently, this is defined as 19, so if TMPDIR=/home/abcdefgh/tmp,
it wouldn't ever be used by programs compiled today. The second part of the
patch affecting info.c changes tmpnam() to tempnam(), which is specified to
allow the user to override the system path. I think this part should be
applied.

-Topi

Message #50 received at 12240@bugs.debian.org (full text, mbox, reply):

From: "¬ü°ê¥J"<usakid@usa.net>

To: "12240@bugs.debian.org" <12240@bugs.debian.org>

Subject: ªü°¶..§Úªº´»°²­p¹º...

Date: Tue, 1 May 2001 23:45:45 +0800

[Message part 1 (text/html, inline)]

Message #55 received at 12240@bugs.debian.org (full text, mbox, reply):

From: ansell168@hongkong.com

To: <1234@www-n3.proxy.aol.com>,<122@news.minx.net.uk>,<12345678@bellsouthips.com>,<1234.567@compuserve.com>,<121803197@pager.icq.com>,<123@www.nudelink.com>,<1230@nab.showmail.net>,<121802877@pager.icq.com>,<12345@67890.com>,<121803083@pager.icq.com>,<121802812@pager.icq.com>,<123@123.com>,<123456@www.fungirls.com>,<12345.abcd1234@isp.com>,<123456@mycollege.edu.uk>,<121803938@pager.icq.com>,<123@shop-oremountain.com>,<122822000921223452780@elcultural.com>,<121803543@pager.icq.com>,<1234@aol.com>,<12240@bugs.debian.org>,<121803824@pager.icq.com>,<12353463e@dwx.com>,<123456@ISP.la.ca.us>,<12305.861193531@hrothgar.gw.com>,<121803340@pager.icq.com>,<121802726@pager.icq.com>,<123@kars.com>,<1234567@mcimail.com>,<121@news.minx.net.uk>,<121803447@pager.icq.com>,<12345@writemail.com>,<121802493@pager.icq.com>

Cc: <12342413156.20010508224029@tanet.net>, <121803827@pager.icq.com>, <121803459@pager.icq.com>, <123@news.minx.net.uk>, <1234567@cityu.edu.hk>, <121803132@pager.icq.com>, <121802835@pager.icq.com>, <1234@maia.proxy.eclipse.net.uk>, <121803649@pager.icq.com>, <121803364@pager.icq.com>, <121802767@pager.icq.com>, <123@123casino.com>, <1234free@onelist.com>, <123456@mycompany.co.uk>, <121802934@pager.icq.com>, <123456@ISP.net>, <1232@foo.bar.com>, <121804130@pager.icq.com>, <121802630@pager.icq.com>, <12306533774.9.SY.KEN@CU20B.COLUMBIA.EDU>, <1234567@skytel.com>, <12345678.90AB@somewhere-else.com>, <123@spider-tj072.proxy.aol.com>, <12396680.989276921513.JavaMail.imail@seamore.excite.com>, <12345@1111.111.11>, <1234@1234.de>, <12285C712E11E.oefr@oawuhtmoadd.org>, <123CyberRegister-unsubscribe@topica.com>, <12345@host212-140-113-49.host.btclick.com>, <122147701@pager.icq.com>, <121803257@pager.icq.com>, <123-insure@lineone.net>, <1226@aol.com>

Subject: life is wonderful today!16555

Date: Sat, 27 Oct 2001 21:36:32 -1700

[Message part 1 (text/html, inline)]

Message #60 received at 12240@bugs.debian.org (full text, mbox, reply):

From: <latest@yahoo.com>

To: <100000-quiet@bugs.debian.org>

Cc: <100000@bugs.debian.org>, <10072@bugs.debian.org>, <12240@bugs.debian.org>, <15753@bugs.debian.org>, <16415@bugs.debian.org>, <17195@bugs.debian.org>, <17624@bugs.debian.org>, <1921@bugs.debian.org>

Subject: Hi !

Date: Thu, 01 Oct 1998 12:59:29 +1200

[Message part 1 (text/plain, inline)]

  FREE   FREE    FREE    FREE    FREE    FREE    FREE    FREE    FREE    


                                       FOR AOL MEMBERS ALSO   

                                 THE GREATEST WEB SITE EVER      

                                          FREE PSYCHIC READING


                                                       Please go to: 

                                                http://66.231.65.36/olp/

                     Please cut and paste this link into your AOL, or internet browser




STILL NOT SURE?  READ BELOW.....




Do you need answers  about life's important questions?  
Questions about love, romance, your spouse?  
Is love, romance, and passion in your future?
Questions about money, finances, your career or job?  
Questions about family matters and life in general? 

We have the answers! Please go to: http://66.231.65.36/olp/
Please cut and paste this link into your AOL browser

Is your spouse cheating or thinking about cheating?  
Is now the time to ask that special someone out on a date?  
Is your current relationship going to get better or worse?  
When should you ask your boss for a raise or a promotion?  
Looking for lucky lottery numbers to win big this week?  

We have the answers! Please go to: http://66.231.65.36/olp/
Please cut and paste this link into your AOLbrowser

Our master psychics have harnessed the power of the internet
to give the most astounding and accurate psychic readings 
for over 4-years now - and we have become the world's leading
online psychic service. Visit us today and you can get a free
horoscope and you can test your psychic ability for free!

Finally, incredible psychic readings available in private over
the internet. Visit us 24 hours a day, 7-days a week, to find
out the answers you need to know to get the most out of your
life -starting right now!  Click here now to find out more 
about our master psychic services.

Please go to: http://66.231.65.36/olp/
Please cut and paste this link into your AOLbrowser

To be removed please go to: http://66.231.65.36/olp/cleanlist.html

Message #65 received at 12240-done@bugs.debian.org (full text, mbox, reply):

From: Jeff Bailey <jbailey@nisa.net>

To: 12240-done@bugs.debian.org

Subject: Very old Bug

Date: Mon, 21 Oct 2002 15:53:03 -0700

This bug was fixed in December of 1998.  Please reopen if you have any
further troubles.

Tks,
Jeff Bailey

-- 
learning from failures is nice in theory...
but in practice, it sucks :)
 - Wolfgang Jaehrling

Send a report that this bug log contains spam.