IRC Log Viewer » #firebreath » 2011-05-27

IRC Nick Time (GMT-7) Message
taxilian 00:05 so generally there is no restriction due to user rights on reading HKLM
appricot 00:05 yes that i know
taxilian 00:05 and in Chrome you are not in a low integrity process (unless something has changed recently), so that should affect it
dougma 00:05 taxilian: back. 2 mins 20 seconds
taxilian 00:05 shouldn't*
appricot 00:05 1
taxilian 00:05 so inside a FireBreath plugin when running Chrome you should have the same rights that you would normally have inside any other process
appricot 00:05 ohh
thanks taxilian
taxilian 00:05 yw
dougma 00:05 yeah so that's quite a bit faster. :)
taxilian 00:05 how much?
dougma 00:05 2 mins 20 seconds
taxilian 00:05 I'm timing mac builds now; just got the PCH working on mac
that's about the same as my mac build was
just over 2 minutes with PCH
I'm timing it without PCH now for comparison
that's… fantastic
you know what stinks about this job? Any time I do anything i have to make it work on all 3 platforms… :-P
so now I just need to make it work on linux
dougma 00:05 dark arts...
taxilian 00:05 heh. this from the guy who does ActiveX stuff? =]
dougma 00:05 :)
taxilian 00:05 FireBreath really has a lot of "dark arts" in it, though =]
so xcode's cache must be *way* better than vs's… there is relatively little difference on mac between using PCH and not
unless I didn't clean completely that time… trying again just in case
dougma 00:05 i just did another rebuild to check
under 3 mins
taxilian 00:05 that takes us finally comparable with mac
which still doesn't seem to care if we use PCH or not
maybe I'll just leave it only enabled on windows
if it doesn't really seem to matter
FireBreathBot 00:05 Commit 1408d65 on master by Richard Bateman: "FIREBREATH-85 Added PCH to PluginCore, NpapiCore"
Commit 4892524 on master by Richard Bateman: "FIREBREATH-85 Added PCH to PluginAuto and AXCore"
Commit ae48965 on master by Richard Bateman: "FIREBREATH-85: PCH works on mac, but doesn't help; disabled"
Commit ff2de58 on master by Richard Bateman: "Merge remote-tracking branch 'origin/master' into taxmaster"
taxilian 00:05 it is in the head now
building on the build server to make sure everythign works on all platforms still
hmm. I think I broke something
appricot 00:05 taxilian : i am trying to open HKLM\Software\Microsoft\Windows\CurrentVersion\Uninstall key with read access
RegOpenKeyEx() fails
taxilian 00:05 have you tested the code you're using in another program to make sure it works?
dougma 00:05 what's GetLastError() return?
taxilian 00:05 another good question =] (dougma can actually probably help you easier than I can)
dougma 00:05 i can try
appricot 00:05 yea tested in an application and worked while given the "UAC Execution Level" to "requireAdministrator" in project properties
taxilian 00:05 what about if you don't?
if UAC is set to normal?
appricot 00:05 GetLastError() returns "Access Denied"
taxilian 00:05 huh. I guess you can't read that particular key as a normal user
dougma 00:05 that's that.
appricot 00:05 I tried many subkeys in HKLM, but result is same
taxilian 00:05 if that's the case (you may want to double check that) your only option is to launch another process and elevate it
appricot 00:05 But no problem with HKCU keys
taxilian 00:05 Perhaps there is another API you could use that would be read specific?
OpenKey is read/write, is it not?
are you specifying KEY_READ? for reference
FireBreathBot 00:05 Commit ea4e4c5 on master by Richard Bateman: "FIREBREATH-85: reverted change needed for mac PCH that broke..."
dougma 00:05 looking at the permissions on the key, Users do have read
taxilian 00:05 wow. I aparently need to go to bed. that's the second commit I've made that either broke or didn't completely fix something
dougma 00:05 'night
taxilian 00:05 gotta fix it first =]
dougma 00:05 mmm yes, borken
taxilian 00:05 think I found it; hang tight
FireBreathBot 00:05 Commit 7c406f4 on master by Richard Bateman: "FIREBREATH-85: Fixed it for real this time"
taxilian 00:05 dougma: that should do you
dougma 00:05 yep, building ok now
taxilian 00:05 dougma: if you wanted to take the time, it wouldn't be hard to add precompiled headers to your own plugin project, which would speed up your build even more
I'm not sure if I want to do it with FBTestPlugin or not because windows PCH is a bit confusing at first (the whole "nothing before this line actually counts" thing)
dougma 00:05 yes, i will look into what you did there
you can really notice the slow down when it gets to FBTestPlugin :)
taxilian 00:05 it's really not hard; create a .cpp and a .h. the .cpp just includes the .h, the .h is the precompiled header
appricot 00:05 taxilian: i am using KEY_READ
taxilian 00:05 then there is a single line that you add before your add_windows_plugin line
appricot: I have no idea, then; perhaps Chrome has started launching plugins in low integrity mode? seems like someone would have noticed sooner, though
dougma: ADD_PRECOMPILED_HEADER(${PROJECT_NAME} "${CMAKE_CURRENT_SOURCE_DIR}/precompiled_headers.h" "${CMAKE_CURRENT_SOURCE_DIR}/precompiled_headers.cpp" SOURCES)
dougma 00:05 appricot: are you using just KEY_READ?
taxilian 00:05 SOURCES is the list of source files that gets passed into add_windows_plugin
appricot: maybe a better question: what is the full line of the command you're using?
appricot 01:05 dougma: yes
taxilian: you mean the RegOpenKeyEx function?
taxilian 01:05 yeah
appricot 01:05 lRet = RegOpenKeyEx(HKEY_LOCAL_MACHINE, pSubKey, 0, KEY_READ, &hKey);
taxilian 01:05 huh
no idea
well, I'm going to bed now
1am here =]
dougma 01:05 bye
_pq_ 01:05 hello
dougma 02:05 hi
dougma 06:05 taxilian_away: ScriptingCore gives some warnings about BOOST_LIB_NAME being redefined. This might be new following PCH?
_pq_ 07:05 going back to xcode 3 as of now
still can't afford 16 gb of ram
taxilian 09:05 dougma: No, thats old
_pq_ 09:05 hello
anyone knows how to add include paths to CMake?
i.e. paths where to search for headers
kalev 09:05 put include_directories(/my/special/header_file_path) in your CMakeLists.txt file
_pq_ 09:05 cool
kalev 09:05 but I would suspect you already have include_directories in there; in that case just add one more entry
_pq_ 09:05 what if i want to include only in a specified platform, since header location differs basing on the OS?
kalev 09:05 one way would be to write a Find module which figures out the path to the headers
but if you are looking for a quick hack, then do something like this:
if (WIN32)
_pq_ 09:05 i added it in X11/projectDef.cmake and it worked
kalev 09:05 set(MYLIB_INCLUDE_DIRS "something")
_pq_ 09:05 yup
kalev 09:05 set(MYLIB_INCLUDE_DIRS "something else")
and then use the variable unconditionally in include_directories() call
ah yes, setting it in X11/projectDef.cmake file is a good idea too, easier than doing conditionals in the base CMakeLists.txt file
sabotaged|wk 14:05 getting this one after merging in the latest from 1.5 branch: make: *** No rule to make target `/tmp/cnt/build/dev/ua/space/apps/firehorn/plugin/build/projects/firehorn/_tmp/Info.plist', needed by `CMakeFiles/cmake.check_cache'. Stop.
anyone have any idea?
j^ 14:05 hi there, anyone got an idea how to best get/set a file in an <input type=file from a plugin, is that possible? passing File/Blob DOM objects around?
sabotaged|wk 15:05 hmm. my build error only happens if there is no build directory, e.g. after a clean checkout
taxilian 15:05 sabotaged|wk: really? hmm. you are the second to have that issue, but the first to identify that particular detail
j^: AFAIK, it is not possible
sabotaged|wk 15:05 clean check out, run prep script once, try a build: fails with that error. run prep script again, and now you can build
can reproduce by just deleting whole build folder and running prep script, then building
taxilian 15:05 interesting
have you tried it with FBTestPlugin? (I'm in the middle of something else, can't try it right now)
if it can be repod with FBTestPlugin I can probably fix it
sabotaged|wk 15:05 i'll try it out
seems to work. i'll have to check diffs between that project and my project's cmake files
taxilian 17:05 sabotaged|wk: if you figure it out, please let me know. I'm rather baffled
sabotaged|wk 17:05 the closest i've gotten is finding out that in the ReRunCMake.make file which is generated on clean checkout, _tmp/Info.plist is listed along with _tmp/InfoPlist.strings and _tmp/Localized.r are lines. but when i re-run the prep script, they disappear and i can build
couldn't find out what the problem was with my projects cmake files
i have a hunch it's this commit, because it was working a while ago on 1.5 for me:
FireBreathBot 17:05 952fb98 by Richard Bateman: Fixed issue where files not always regenerated when needed