Attachment 18322 Details for Bug 13213 – patches for 4.17

[patch] patches for 4.17

samba-repo-build-4.17.patch (text/plain), 12.02 KB, created by Douglas Bagnall on 2024-06-05 03:04:40 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Douglas Bagnall

Created: 2024-06-05 03:04:40 UTC

Size: 12.02 KB

patch

obsolete

>From 155dacc34bc8f089a068511ef69245c969c43559 Mon Sep 17 00:00:00 2001
>From: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Date: Wed, 29 May 2024 12:14:04 +1200
>Subject: [PATCH 1/4] pidl:Typelist: resolveType(): don't mistake a reference
> for a name
>
>This function is only used by Python.pm, and was assuming any argument
>unrecognised by hasType is a name. It sometimes isn't, resulting in
>structures like this:
>
>	{
>	  'DATA' => {
>		      'TYPE' => 'STRUCT'
>		    },
>	  'NAME' => {
>		      'TYPE' => 'STRUCT',
>		      'ALIGN' => undef,
>		      'SURROUNDING_ELEMENT' => undef,
>		      'ORIGINAL' => {
>				      'TYPE' => 'STRUCT',
>				      'FILE' => 'source3/librpc/idl/smbXsrv.idl',
>				      'LINE' => 101,
>				      'NAME' => 'tevent_context'
>				    },
>		      'ELEMENTS' => undef,
>		      'NAME' => 'tevent_context',
>		      'PROPERTIES' => undef
>		    },
>	  'TYPE' => 'TYPEDEF'
>      };
>
>The problem with that is we end up with the HASH reference as a name
>in Python bindings, like this
>
>      PyErr_SetString(PyExc_TypeError, "Can not convert C Type struct HASH(0x5e2dfe5ee278) from Python");
>
>which makes the build nondeterministic (as well as making the message
>a little mysterious).
>
>I think all the structures for which this happens are marked
>'[ignore]' in IDL, meaning they are not transmitted on the wire. They
>should perhaps also not have useless Python getsetters, but let's call
>that a different problem.
>
>Thanks to Freexian and the Debian LTS project for sponsoring this work.
>
>BUG: https://bugzilla.samba.org/show_bug.cgi?id=13213
>
>Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Reviewed-by: Andrew Bartlett <abartlet@samba.org>
>(cherry picked from commit f3433f60b8ee83fc785a6e6838513de31bff5a6a)
>---
> pidl/lib/Parse/Pidl/Typelist.pm | 14 ++++++++++++--
> 1 file changed, 12 insertions(+), 2 deletions(-)
>
>diff --git a/pidl/lib/Parse/Pidl/Typelist.pm b/pidl/lib/Parse/Pidl/Typelist.pm
>index 2b850ffee4b..c37d6114bd1 100644
>--- a/pidl/lib/Parse/Pidl/Typelist.pm
>+++ b/pidl/lib/Parse/Pidl/Typelist.pm
>@@ -120,8 +120,18 @@ sub resolveType($)
> 	my ($ctype) = @_;
> 
> 	if (not hasType($ctype)) {
>-		# assume struct typedef
>-		return { TYPE => "TYPEDEF", NAME => $ctype, DATA => { TYPE => "STRUCT" } };
>+		if (! ref $ctype) {
>+			# it looks like a name.
>+			# assume struct typedef
>+			return { TYPE => "TYPEDEF", NAME => $ctype, DATA => { TYPE => "STRUCT" } };
>+		}
>+		if ($ctype->{NAME} && ($ctype->{TYPE} eq "STRUCT")) {
>+			return {
>+				TYPE => "TYPEDEF",
>+				NAME => $ctype->{NAME},
>+				DATA => $ctype
>+			};
>+		}
> 	} else {
> 		return getType($ctype);
> 	}
>-- 
>2.34.1
>
>
>From 23e312ff284fef9f243cce5dacb7b36d26f7e4c8 Mon Sep 17 00:00:00 2001
>From: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Date: Tue, 28 May 2024 12:17:06 +1200
>Subject: [PATCH 2/4] buildtools:pidl: avoid hash randomisation in pidl
>
>Like many languages, Perl uses has randomisation to prevent nasty
>users using crafted values that hash to the same number to effect a
>denial of service. This means the traversal order of perl HASH tables
>is different every time.
>
>The IDL handed to pidl is trusted, so we don't really need
>randomisation, but we do want to be sure the build is the same every
>time.
>
>I am not aware of hash randomisation causing problems, but it seems
>prudent to avoid it.
>
>We do a similar thing with PYTHONHASHSEED for the entire build.
>
>Thanks to Freexian and the Debian LTS project for sponsoring this work.
>
>BUG: https://bugzilla.samba.org/show_bug.cgi?id=13213
>
>Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Reviewed-by: Andrew Bartlett <abartlet@samba.org>
>(cherry picked from commit 068b366709d005275727a0e8929d272c04cb7bd8)
>---
> buildtools/wafsamba/samba_pidl.py | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
>diff --git a/buildtools/wafsamba/samba_pidl.py b/buildtools/wafsamba/samba_pidl.py
>index 0ce860ca6ae..2b079392870 100644
>--- a/buildtools/wafsamba/samba_pidl.py
>+++ b/buildtools/wafsamba/samba_pidl.py
>@@ -81,7 +81,9 @@ def SAMBA_PIDL(bld, pname, source,
>         else:
>             cc = 'CC="%s"' % bld.CONFIG_GET("CC")
> 
>-    t = bld(rule='cd ${PIDL_LAUNCH_DIR} && %s%s %s ${PERL} ${PIDL} --quiet ${OPTIONS} --outputdir ${OUTPUTDIR} -- "${IDLSRC}"' % (pidl_dev, cpp, cc),
>+    t = bld(rule=('cd ${PIDL_LAUNCH_DIR} && PERL_HASH_SEED=0 %s%s %s ${PERL} '
>+                  '${PIDL} --quiet ${OPTIONS} --outputdir ${OUTPUTDIR} -- "${IDLSRC}"' %
>+                  (pidl_dev, cpp, cc)),
>             ext_out    = '.c',
>             before     = 'c',
>             update_outputs = True,
>-- 
>2.34.1
>
>
>From ec2b3e5515f259d1cef4f0897685a0f0d414abfc Mon Sep 17 00:00:00 2001
>From: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Date: Tue, 28 May 2024 11:34:51 +1200
>Subject: [PATCH 3/4] examples:winexe: reproducible builds with zero timestamp
>
>Windows Portable Executable files have a timestamp field and a
>checksum field. By default the timestamp field is updated to the
>current time, which consequently changes the checksum. This makes the
>build nondeterministic. It looks like this:
>
>  --- a/tmp/winexe-1/winexesvc64_exe_binary.c
>  +++ b/tmp/winexe-2/winexesvc64_exe_binary.c
>  @@ -23,7 +23,7 @@ const DATA_BLOB *winexesvc64_exe_binary(void)
>                0x6D, 0x6F, 0x64, 0x65, 0x2E, 0x0D, 0x0D, 0x0A,
>                0x24, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
>                0x50, 0x45, 0x00, 0x00, 0x64, 0x86, 0x0A, 0x00,
>  -             0xB2, 0x16, 0x55, 0x66, 0x00, 0x00, 0x00, 0x00,
>  +             0xD3, 0x3B, 0x55, 0x66, 0x00, 0x00, 0x00, 0x00,
>                0x00, 0x00, 0x00, 0x00, 0xF0, 0x00, 0x2E, 0x02,
>                0x0B, 0x02, 0x02, 0x26, 0x00, 0x86, 0x00, 0x00,
>                0x00, 0xBA, 0x00, 0x00, 0x00, 0x0C, 0x00, 0x00,
>  @@ -33,7 +33,7 @@ const DATA_BLOB *winexesvc64_exe_binary(void)
>                0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
>                0x05, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00,
>                0x00, 0x40, 0x01, 0x00, 0x00, 0x04, 0x00, 0x00,
>  -             0x73, 0xD7, 0x00, 0x00, 0x03, 0x00, 0x60, 0x01,
>  +             0x94, 0xFC, 0x00, 0x00, 0x03, 0x00, 0x60, 0x01,
>                0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x00, 0x00,
>                0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
>                0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00,
>
>https://learn.microsoft.com/en-us/windows/win32/debug/pe-format says
>that a timestamp of zero can be used to represent a time that is not
>"real or meaningful", so we do that.
>
>As far as I can tell, the timestamp and checksum are only used in
>DLLs, not directly executed .exe files.
>
>Thanks to Freexian and the Debian LTS project for sponsoring this work.
>
>BUG: https://bugzilla.samba.org/show_bug.cgi?id=13213
>
>Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Reviewed-by: Andrew Bartlett <abartlet@samba.org>
>(backported from commit e604f7575167d3572e1b67c6e77ab7273508533d)
>[dbagnall@samba.org: adapted for wscript_build differences]
>---
> examples/winexe/wscript       | 14 ++++++++++++++
> examples/winexe/wscript_build |  4 ++--
> 2 files changed, 16 insertions(+), 2 deletions(-)
>
>diff --git a/examples/winexe/wscript b/examples/winexe/wscript
>index 6b311b1da41..a202282e765 100644
>--- a/examples/winexe/wscript
>+++ b/examples/winexe/wscript
>@@ -1,4 +1,6 @@
> #!/usr/bin/env python
>+import os
>+
> 
> def configure(conf):
>     AR32 = ['i386', 'i586', 'i686']
>@@ -27,5 +29,17 @@ def configure(conf):
>             conf.DEFINE('HAVE_WINEXE_CC_WIN64', 1);
>             break
> 
>+    source_date_epoch = os.environ.get('SOURCE_DATE_EPOCH')
>+    if source_date_epoch is None:
>+        # We set the timestamp to be 0, which makes the winexe build
>+        # reproducible. According to
>+        # https://learn.microsoft.com/en-us/windows/win32/debug/pe-format
>+        #
>+        # > If the stamp value is 0 or 0xFFFFFFFF, it does not
>+        # > represent a real or meaningful date/time stamp.
>+
>+        source_date_epoch = '0'
>+
>+    conf.env.SOURCE_DATE_EPOCH = source_date_epoch
>     conf.DEFINE("WINEXE_LDFLAGS",
>                 "-s -Wall -Wl,-Bstatic -Wl,-Bdynamic -luserenv")
>diff --git a/examples/winexe/wscript_build b/examples/winexe/wscript_build
>index 143739f3de0..bb454159e08 100644
>--- a/examples/winexe/wscript_build
>+++ b/examples/winexe/wscript_build
>@@ -70,7 +70,7 @@ if bld.env.WINEXE_CC_WIN32:
>         'winexesvc32_exe',
>         source='winexesvc.c',
>         target='winexesvc32.exe',
>-        rule='${WINEXE_CC_WIN32} ${SRC} -o ${TGT} ${WINEXE_LDFLAGS}')
>+        rule='SOURCE_DATE_EPOCH=${SOURCE_DATE_EPOCH} ${WINEXE_CC_WIN32} ${SRC} -o ${TGT} ${WINEXE_LDFLAGS}')
>     vars = {"WINEXE_FN": "winexesvc32_exe_binary"}
>     bld.SAMBA_GENERATOR(
>         'winexesvc32_exe_binary',
>@@ -86,7 +86,7 @@ if bld.env.WINEXE_CC_WIN64:
>         'winexesvc64_exe',
>         source='winexesvc.c',
>         target='winexesvc64.exe',
>-        rule='${WINEXE_CC_WIN64} ${SRC} -o ${TGT} ${WINEXE_LDFLAGS}')
>+        rule='SOURCE_DATE_EPOCH=${SOURCE_DATE_EPOCH} ${WINEXE_CC_WIN64} ${SRC} -o ${TGT} ${WINEXE_LDFLAGS}')
>     vars = {"WINEXE_FN": "winexesvc64_exe_binary"}
>     bld.SAMBA_GENERATOR(
>         'winexesvc64_exe_binary',
>-- 
>2.34.1
>
>
>From 86d249275e18224f7d3367f7fc0d2bd93c659ec2 Mon Sep 17 00:00:00 2001
>From: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Date: Thu, 30 May 2024 10:44:24 +1200
>Subject: [PATCH 4/4] examples:winexe: embed Samba version as exe timestamp
>
>It turns out the timestamp doesn't need to be real, and it isn't used,
>but it might as well tell you something. So let's make it tell you what
>version of Samba it came from, which could be useful for people who have
>lots of old winexes lying around, the poor souls.
>
>00000040  0e 1f ba 0e 00 b4 09 cd  21 b8 01 4c cd 21 54 68  |........!..L.!Th|
>00000050  69 73 20 70 72 6f 67 72  61 6d 20 63 61 6e 6e 6f  |is program canno|
>00000060  74 20 62 65 20 72 75 6e  20 69 6e 20 44 4f 53 20  |t be run in DOS |
>00000070  6d 6f 64 65 2e 0d 0d 0a  24 00 00 00 00 00 00 00  |mode....$.......|
>00000080  50 45 00 00 64 86 0a 00  00 15 04 00 00 00 00 00  |PE..d...........|
>                                    |  |  |
>                                    |  |  major  4.
>                                    |  minor     21.
>                                    release      0
>
>BUG: https://bugzilla.samba.org/show_bug.cgi?id=13213
>
>Signed-off-by: Douglas Bagnall <douglas.bagnall@catalyst.net.nz>
>Reviewed-by: Andrew Bartlett <abartlet@samba.org>
>
>Autobuild-User(master): Douglas Bagnall <dbagnall@samba.org>
>Autobuild-Date(master): Fri May 31 01:28:06 UTC 2024 on atb-devel-224
>
>(cherry picked from commit 3a7dbf8b77b2a9e7cdc55bc5b339b9f501d037aa)
>---
> examples/winexe/wscript | 21 ++++++++++++++-------
> 1 file changed, 14 insertions(+), 7 deletions(-)
>
>diff --git a/examples/winexe/wscript b/examples/winexe/wscript
>index a202282e765..c4f13b89f01 100644
>--- a/examples/winexe/wscript
>+++ b/examples/winexe/wscript
>@@ -31,14 +31,21 @@ def configure(conf):
> 
>     source_date_epoch = os.environ.get('SOURCE_DATE_EPOCH')
>     if source_date_epoch is None:
>-        # We set the timestamp to be 0, which makes the winexe build
>-        # reproducible. According to
>-        # https://learn.microsoft.com/en-us/windows/win32/debug/pe-format
>+        # We use the version to make up the timestamp that will be
>+        # embedded in winexe.exe, to keep the build reproducible.
>         #
>-        # > If the stamp value is 0 or 0xFFFFFFFF, it does not
>-        # > represent a real or meaningful date/time stamp.
>-
>-        source_date_epoch = '0'
>+        # This is less evil than it sounds. According to Raymond Chen in
>+        # https://devblogs.microsoft.com/oldnewthing/20180103-00/?p=97705
>+        # since Windows 10 the timestamp has been randomised.
>+        #
>+        # The purpose of the timestamp in Windows PE files seems to be
>+        # to make spotting ABI changes in DLLs quicker, for which a
>+        # random number is just as good as a real time. The timestamp
>+        # in .exe files is not used.
>+        import samba_version
>+        v = samba_version.load_version(env=conf.env)
>+        version = (v.MAJOR << 16) | (v.MINOR << 8) | v.RELEASE
>+        source_date_epoch = str(version)
> 
>     conf.env.SOURCE_DATE_EPOCH = source_date_epoch
>     conf.DEFINE("WINEXE_LDFLAGS",
>-- 
>2.34.1
>

Actions: View

Attachments on bug 13213: 18318 | 18321 | 18322