#80 nicer registry commands log

[Instructor]

install.log (without patch):

WriteRegStr: set -2147483647\SOFTWARE\Opera
Software\a\X to y
WriteRegStr: set -2147483647\SOFTWARE\Opera
Software\a\\X\/\ to y
WriteRegDWORD: set -2147483647\SOFTWARE\Opera
Software\a\X2 to 12345
WriteRegBin: set -2147483647\SOFTWARE\Opera
Software\a\X3 with 3 bytes
DeleteRegValue: -2147483647\SOFTWARE\Opera
Software\a\X2
DeleteRegKey: -2147483647\Software\Opera Software\a
RMDir: RemoveDirectory invalid input("")

install.log (with patch):

WriteRegStr:
"HKEY_CURRENT_USER\SOFTWARE\Opera
Software\a" "X"="y"
WriteRegExpandStr:
"HKEY_CURRENT_USER\SOFTWARE\Opera
Software\a" "\X\/\"="y"
WriteRegDWORD:
"HKEY_CURRENT_USER\SOFTWARE\Opera
Software\a" "X2"="12345"
WriteRegBin:
"HKEY_CURRENT_USER\SOFTWARE\Opera
Software\a" "X3"="01ffdd"
DeleteRegValue:
"HKEY_CURRENT_USER\SOFTWARE\Opera
Software\a" "X2"
DeleteRegKey:
"HKEY_CURRENT_USER\Software\Opera Software\a"
RMDir: RemoveDirectory invalid input("")

[Amir Szekely]

Logged In: YES
user_id=584402

You forgot to attach the patch. Make sure you check the box,
it's always forgotten.

[Instructor]

Ou, ofcourse checkbox..

[Amir Szekely]

Logged In: YES
user_id=584402

Thanks.

Your patch has effects, even when not compiled with
NSIS_CONFIG_LOG.

1) In EW_WRITEREG, you can use an if block on rtype when
type is REG_SZ instead of checking if type is REG_SZ_EXPAND.
2) Why remove `// use buf2, buf3 and buf4`? It's still true.
3) Only WriteRegBin uses buf4, don't define it. Use bufs[4].

And also, there're some other problems:

4) GetCompressedDataFromDataBlockToMemory changes buf3 when
using WriteRegBin. It also uses buf4, which is ran over by
my_HexDataToHexStr.
5) Please rename the new functions to _FuncName and the
define to the real function name. The my_ prefix is usually
used in the code for API wrappers.
6) buf4 is not large enough to contain the hex
representation of 3072 bytes, let alone 1024 bytes. I'm also
not so sure the entire block should be written to the log.
The first few bytes should be enough.
4) Advancing the pointer in HexDataToHexStr would be more
efficient than using lstrcat.

[Instructor]

Logged In: YES
user_id=1197108

4) I have some problem with what functions what buffers
changes, because I don't see any problem with correct log
output.
6) I have thought that user can't input more then
NSIS_MAX_STRLEN characters (AABBCC...). Now if number
of bytes will be more then 8 it will be trimmed like
"AABBCCDDAABBCCDD..."
7) As it only 8 bytes to save to log probably lstrcat will be ok

[Instructor]

[Amir Szekely]

Logged In: YES
user_id=584402

4) Have you used a long enough binary string? You need at
least 2048 bytes of data for buf4 to be used.
6) Since version 2.01, WriteRegBin accepts 3 *
NSIS_MAX_STRLEN. You can see it in the function call where 3
* NSIS_MAX_STRLEN is passed.
7) Indeed.

[Instructor]

Logged In: YES
user_id=1197108

Simply used another buffer bufs[5] instead buf3.

[Instructor]

[Amir Szekely]

Logged In: YES
user_id=584402

There's no such thing as bufs[5], and it still uses bufs[4]
which can be overridden. Maybe you can call the
uncompression function twice. Once with just a little data
for the log, and then again for the entire data block. Make
sure it works with all compression types.

[Instructor]

Logged In: YES
user_id=1197108

Can you explain what do the line:
size = GetCompressedDataFromDataBlockToMemory(parm3,
data, 3 * NSIS_MAX_STRLEN);

and what buffers it changes

[Amir Szekely]

Logged In: YES
user_id=584402

It uncompresses data pointed by parm3 into data with a
maximum of 3 * NSIS_MAX_STRLEN bytes uncompressed. All you
really need to know about it, is what the comment above it
says - it affects buf2, buf3 and buf4.

[Instructor]

[Instructor]

Logged In: YES
user_id=1197108

Another try, using rootstr[24]

[Instructor]

Logged In: YES
user_id=1197108

Any comments? I hope buf3 not used after
GetCompressedDataFromDataBlockToMemory

[Amir Szekely]

Logged In: YES
user_id=584402

The same still goes for bufs[3]. You can't use it after
GetCompressedDataFromDataBlockToMemory.

You also haven't address issue #1 in my original comment.
With your patch, the code size will grow even without
NSIS_CONFIG_LOG.

[Instructor]

Logged In: YES
user_id=1197108

Hope the last variant (double calling
GetCompressedDataFromDataBlockToMemory and
RegKeyHandleToName)

[Instructor]

[Amir Szekely]

Logged In: YES
user_id=584402

That should work. Have you tested it using all compression
types?

Why put the hex data in buf3 and the root key name in buf2
again? Why not simply putting the hex data in buf2?

You still haven't address issue number 1 in my original
comment. Your code still affects the installer, even without
NSIS_CONFIG_LOG.

[Instructor]

Logged In: YES
user_id=1197108

>Have you tested it using all compression
types?
Yes

>Why put the hex data in buf3 and the root key name in buf2
>again? Why not simply putting the hex data in buf2?
1. "LPBYTE data = (LPBYTE) buf2;" so it will be the same
buffer "LogHexData(parm3, data, buf2, 8);"
2. As function "LogHexData" use
"GetCompressedDataFromDataBlockToMemory", it can
change buf3, therefore function "RegKeyHandleToName" uses
twice.

>You still haven't address issue number 1 in my original
>comment.
Sorry that I ignore the previous the same comment, I simply
don't understand what do you mean.

[Amir Szekely]

Logged In: YES
user_id=584402

I've modified and applied the patch. The modified patch is
attached. Thanks for the patch.

If you want, drop by the irc channel and I'll try to better
explain what I wanted in issue #1.

[Amir Szekely]

the applied patch