689f6a5c62
bug #75504 ; sr=jband. making it impossible to misuse |getter_AddRefs| in the way jband discovered
2001-05-03 10:15:09 +00:00
1be564909a
Checkin shaver's patch to allow debug Mozilla builds to dynamicly link with XPCOM-based, nsCOMPtr-using optimized plugins (77112, r=brendan, sr=scc, a=chofmann).
2001-04-24 21:13:15 +00:00
093e897c1b
Fix bustage on VC++ and gcc 2.7.2.x caused by incorrect overload resolution.
2001-04-07 17:25:16 +00:00
244ee9e63a
Replace my previous incorrect fix for |operator==| ambiguities with a correct one: provide additional |operator==| and |operator!=| for comparing nsCOMPtrs to raw pointers that take a non-const raw pointer argument so that builtin operators will not have a better conversion for one argument. b=65664 r=waterson@netscape.com sr=scc@mozilla.org
2001-04-07 15:29:39 +00:00
8d34652545
Make |nsCOMPtr<T>::operator&| private so that it is not used by clients. Legitimate users should use |address_of| instead. b=59414 r=disttsc@bart.nl sr=scc@mozilla.org
...
Fix ambiguities relating to the various forms of |operator==| defined for nsCOMPtrs by avoiding |const| for parameters for compilers where it (incorrectly) confuses overload resolution. r=scc@mozilla.org sr=brendan@mozilla.org b=65664
2001-02-04 19:57:29 +00:00
198833687a
Attempt to fix |address_of| bustage on IRIX. r=scc@mozilla.org b=59414
2000-12-10 03:13:53 +00:00
7f9852c286
Work in progress for bug 59414. Prepare to make |nsCOMPtr|'s |operator&| private to prevent many cases of accidental usage, and provide |address_of| for clients who really want a pointer to an nsCOMPtr. r=scc@mozilla.org sr=waterson@netcape.com
2000-12-09 04:45:43 +00:00
847d576a56
Workaround for long-standing AIX bustage
2000-07-28 08:11:51 +00:00
b3e435e1de
# 37239
...
r = valeski, a = brendan
OS/2 bring-up - Yes I know this is ugly.
We are getting a linker error on this delete.
We want to get our builds up by any means necessary,
so I need to comment this out for us for now.
2000-06-14 00:26:52 +00:00
ab3cce39a1
further logging fixes; removed |explicit| to "nscore.h"; removed old-style logging definitions for |NSCAP_ADDREF| and |NSCAP_RELEASE|, define them externally if you want them (the new comment says how); fixed some comments... I'll be moving to doc-style comments soon to better support doxygen
2000-06-10 09:58:13 +00:00
42a6c98a0f
further fix for the defined case. r=dbaron
2000-06-08 05:08:14 +00:00
129f2bc511
Improved logging facilities on |nsCOMPtr| in cooperation with dbaron, for his new logging tools. r=dbaron
2000-06-08 04:52:39 +00:00
7ea63b347d
Canonify HAVE_CPP_ACCESS_CHANGING_USING and HAVE_CPP_AMBIGUITY_RESOLVING_USING.
2000-06-05 00:46:37 +00:00
a1285d663d
Bug #39824 . Checking in patch from dp. This patch fixes exports on some |nsCOMPtr_helper| classes so we can link appropriately. r=scc.
2000-05-20 18:31:45 +00:00
ec50330ca2
Remove #ifdef that is unnecessary when STL headers are not included.
2000-05-14 00:05:24 +00:00
6d1bd93d73
Add necessary code to get NEW_STRING_APIS to compile properly on gcc-2.7.2.3. r=scc
2000-05-13 00:05:29 +00:00
d9a28087cd
changed some comments
2000-05-12 05:20:21 +00:00
e2976c60e9
Beef up autoconf tests to differentiate 'ambiguity resolving' and 'access changing' support of the 'using' keyword. Add a test to see if the 'std' namespace is used on the platform. r=scc
2000-03-28 07:01:37 +00:00
06d7fc7dff
slight renaming and fixed a template signature for AIX
2000-03-22 22:52:17 +00:00
1099be693b
Added the Bill Law trick to make |nsCOMPtr| break the build for everybody where it use to break it only for Solaris and a few others (the `tainting' problem). Added a haiku. r=danm
2000-03-18 05:17:46 +00:00
ba8a0f115a
Renamed an internal typename. Was |nsDontAddRef<T>|, now is |already_AddRefed<T>| in hopes that it can be used outside of "nsCOMPtr.h". Moved |HAVE_CPP_...| symbols out of here and into "nscore.h" because more files need them. r=pinkerton
2000-03-17 23:08:18 +00:00
5b719b7bc5
wrapped |nsCOMPtr| vs. raw pointer comparisons in a |#ifndef| for |NSCAP_NSCOMPTR_TO_RAW_COMPARISONS_ARE_AMBIGUOUS| which is currently not defined. When an autoconf test defines this symbol, these operators will go away, and Linux will never be bit by the |nsCOMPtr| can't compare to a raw pointer error. Currently, there is no effect. r={mjudge,rginda}
2000-02-12 23:30:52 +00:00
2471a0d722
enable the |ASSERT_NO_QUERY_NEEDED| tests for the |getter_AddRefs()| case as well. bug #26420 , r=dougt
2000-02-06 21:17:51 +00:00
d220bcf442
Fixed tabs, some comments, etc. Made |do_QueryInterface()| not participate in the |...TEST_DONTQUERY_CASES| test to help us debug bug #25327 . r=pinkerton
2000-01-29 22:29:08 +00:00
fc8ef76eff
macros and defs to removed conditional definition of |GetIID()| within |nsISupports|. r=beard, (waterson?), bug=10477
2000-01-13 05:48:47 +00:00
7f1ef3261d
Fix specialized |CallQueryInterface| as per review by travis ... r=travis.
1999-12-08 01:44:06 +00:00
ccc2279daa
Adding an alternate version of |CallQueryInterface| as an experiment on behalf of travis. This function is not yet called, and should not effect the build or the runtime. r=hyatt
1999-12-07 14:09:29 +00:00
290a51fcd4
fixed an annoying, but rare, bug where an |nsCOMPtr| in an ownership ring could |Release()| twice. See the comment in the code for details. r=waterson
1999-11-20 08:19:24 +00:00
913a64bdee
missed two T-->nsISupports instances, which primary platforms didn't catch, since they weren't instantiated
1999-11-08 12:25:11 +00:00
50bb3ffae5
Specialize |nsCOMPtr| for |nsISupports| (i.e., |nsCOMPtr<nsISupports>|). The specialization _does_not_ have the invariant that |nsCOMPtr<T>| has of always pointing to the COM-correct interface. An |nsCOMPtr<nsISupports>| essentially means "a pointer to _any_ [XP]COM interface". This change allows dougt to fix his M11 bug. a=chofmann, r={danm, hyatt}
1999-11-08 11:43:11 +00:00
8535dda53e
updated xPL license boilerplate to v1.1, a=chofmann@netscape.com,r=endico@mozilla.org
1999-11-06 03:43:54 +00:00
e941d5a661
backing out changes. looks like this stuff is untested in a full build.
1999-11-05 03:31:04 +00:00
fa6602724f
Fix for the template variable change from <T> to <nsISupports>. r=dougt
1999-11-05 03:11:03 +00:00
777d0f9159
Fix for 17842. Make nsCOMPtr<nsISupports> should not be weak typed. scc's
...
this change to me and since he is on vacation until monday and I need this
for an M11 bug, I am checking it in. I am hoping it is the right
thing to do.
a=brendan.
1999-11-05 02:18:46 +00:00
2571930b5e
shutting up copious warnings using /W4 on MSVC
1999-11-03 07:39:19 +00:00
8f4e7ad8e4
changed the mechanism underlying |do_QueryInterface| to decouple |nsCOMPtr| from services, |nsIWeakReference|, etc. This change fixes bug #17364 , and paves the way for making |nsCOMPtr| cooperate with services. r=valeski (additional builds done by dmose and syd)
1999-10-31 00:35:48 +00:00
cffa6a515e
some small documentation changes, plus removed a function that was otherwise unused, and was confusing Solaris.
1999-08-26 10:45:50 +00:00
06e8699a24
fixed tabs
1999-08-23 10:07:16 +00:00
944678da9d
renamed to QueryReferent, do_QueryReferent, etc.
1999-08-13 23:00:58 +00:00
1f0ab10430
experiment: I don't think it like the forward declaration
1999-08-03 10:44:12 +00:00
e4d1a4321f
Moved the typedef to its own file: "nsWeakPtr.h". This avoids the circular includes problem and lets people #include just "nsWeakPtr.h" when they want an |nsWeakPtr|.
1999-08-03 08:15:31 +00:00
b5ee2a318a
OK, now only nsCOMPtr.h includes nsIWeakReference, and not the other way around. I'll probably have to make a header file just to contain the one-line typedef I moved.
1999-08-03 08:10:08 +00:00
eaeff321cf
Well, not everybody could do it just from the forward declared class. Yes, nsCOMPtr.h and nsIWeakReference.h include eachother, but it works anyway (they test for already being included).
1999-08-03 07:44:53 +00:00
dfa83459be
Added |do_QueryReference| for initializing/assigning from |nsWeakPtr|s. Also some fixes.
1999-08-03 07:21:39 +00:00
73a836eeaa
Allow explicit disabling of |...TEST_DONTQUERY_CASES| and |...DEBUG_PTR_TYPES| on the command-line for problematic compilers like QNX et al. This allows me to fix the QNX build, and provide an answer to some people compiling with the Sun Workshop.
1999-07-22 21:09:49 +00:00
4adb195c6c
switching back to the stricter declaration of |nsDerivedSafe<T>::operator=()|. Since I removed |nsISupportsArray::operator=()|, all should be well. I left the old declaration in but commented out, just in case. I hope I don't have to conditionalize this wth an autoconf test. I've built this (with the help of several other people including mjudge, waqar, slamm, et al) on as many other platforms as I could. This declaration is intended to stop people from declaring an |operator=()| in their (XP)COM interface classes.
1999-07-22 20:03:42 +00:00
9d49cb3127
new feature lets you `see past' |nsISupports| in |nsCOMPtr|s. Now in the debugger (in debug versions only) you will see the actual type of the object. This is enabled by defining the preprocessor symbol |NSCAP_FEATURE_DEBUG_PTR_TYPES|. It cannot be turned on in non-debug builds. It is currently on in debug builds by default. It may become a feature that has to be enabled with a build command line switch.
NSCAP_FEATURE_ALLOW_RAW_POINTERS is now always the case and no longer optional; NSCAP_FEATURE_ALLOW_COMPARISONS is now always the case and no longer optional; cleaned up the previous definition and invocations of the |Assert_NoQueryNeeded| test.
1999-07-15 03:59:09 +00:00
cc711499f4
quick fix for bug 9250 - assert on startup
1999-06-30 07:13:26 +00:00
f8f7778a8e
partial backout of scc's nsCOMPtr, attempting to revive Win32 build
1999-06-29 22:33:18 +00:00
d10f87cd9d
added logic to test whether QueryInterface was actually required when it is ommitted, |nsISupports::GetIID()| --> |nsCOMTypeInfo<nsISupports>::GetIID()|, changed the scheme by which we compare with zero to a more robust mechanism, modified specialized |CallQueryInterface| to use the global version
1999-06-29 21:09:59 +00:00
0c87834eef
fixed for VMS
1999-06-03 20:36:53 +00:00
16cc6edf13
replaced tabs with spaces as per our standards
1999-06-02 01:59:28 +00:00
5c2933b1c2
added comparisons that will allow |nsnull != myPtr| et al
1999-06-02 01:29:55 +00:00
5f7f07adc9
Fixup CallQueryInterface
1999-05-12 04:50:08 +00:00
4fb41627b5
temporarily commented out CallQueryInterface for UNIX, added NS_EXPORT for Windows
1999-05-10 21:43:29 +00:00
64fea0e7f5
conditionally factored destructor
1999-05-10 20:48:43 +00:00
4c26313504
sense of HAVE_CPP_BOOL was backwards, fixed it (thanks, Jim Dunn)
1999-05-05 23:36:38 +00:00
6c7fa04eb5
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Wow! Re-enabled initialization/assignment from raw COM interface pointers, only this time it means the same as |dont_QueryInterface|. Which is what people want and expect. Re-enabled comparisons between |nsCOMPtr|s and raw COM interface pointers; and it means what people expect: are these two 4-byte values equal/not-equal. Added a function for comparing two pointers (either can be raw or nsCOMPtr) called |SameCOMIdentity|, which provides the other kind of comparison.
No current functionality has changed. Everything you're doing now is still legal and valid. These changes only bring some additional options.
-----BEGIN PGP SIGNATURE-----
Version: PGP Personal Privacy 6.0.2
Comment: get my key at < http://www.meer.net/ScottCollins/#key >
iQA/AwUBNzAeufGmojMuVn+fEQInWgCfUVK/7xfd/pdhAmyIe4HhynKsuucAn20t
MRtmNY4e2MAdwClvhR65a5Yv
=elb3
-----END PGP SIGNATURE-----
1999-05-05 10:37:05 +00:00
314436f96d
change |CallQueryInterface| to handle situations where |nsISupports| is ambiguous.
1999-05-04 12:40:52 +00:00
f5a32343a4
added |null_nsCOMPtr()|, |CallQueryInterface()|, fixed comments; added |const| in several appropriate places
1999-05-04 09:37:35 +00:00
8eb511fa4b
inlining the work of |StartAssigment| reduces |nsCOMPtr| overhead on windows.
1999-03-23 20:35:49 +00:00
31ff626ae9
rename |IID()| --> |GetIID()|
1999-03-03 19:48:57 +00:00
9d21ae3942
backed out depency on nscore.h
1999-03-02 23:52:38 +00:00
8058bf1464
now rely on "nscore.h" for NS_REINTERPRET_CAST instead of defining my own. now relying on autoconf system for all unixes. Some tests may need to be fixed.
1999-03-02 23:08:54 +00:00
49754e4257
Adding comparisons to autoconf results.
1999-03-01 10:13:18 +00:00
0875a55f81
AIX portability fixes. (Waqar Malik <waqar@netscape.com>)
1999-03-01 06:29:14 +00:00
312b8d76d7
for VC++5.0, no member using declarations (to prevent internal compiler error); to make this compile fixed type of AddRef and Release in nsDerivedSafe
1999-02-26 23:43:40 +00:00
0238953fe3
deprecated |dont_AddRef|, added a new form of |getter_AddRefs| to take its place. Removed all nothing of bool from the headers (should help gcc)
1999-02-26 19:50:46 +00:00
e7831871be
fix AIX bustage
1999-02-26 18:58:08 +00:00
643dbd9b6d
fix HPUX bustage - HP's compiler can't do "using" or static casts
1999-02-25 17:47:54 +00:00
99b1533740
Adding support for Solaris native compiler. a=scc
1999-02-18 22:34:15 +00:00
d4ae494464
updated to new interface for nsCOMPtr
1999-02-17 01:58:25 +00:00
f020767dfc
Added |nsresult nsCOMPtr::assignment_error() const;| which returns the error, if any, from the most recent operation (constructor or assignment operator) that implicitly calls |QueryInterface|. This did not increase the size of an |nsCOMPtr|.
1999-02-11 01:34:07 +00:00
edaa2f7db9
Made the constructor explicit again (see comment) to stop people from breaking the Linux builds.
1999-01-22 01:17:48 +00:00
0c7a4ca128
OK. We were wrong about the SOLARIS case. Sorry.
1999-01-18 20:32:41 +00:00
be87c1b10a
Added patch for bsdi from netscape.public.mozilla.patches; removed test for !SOLARIS as agreed with macafee.
1999-01-18 18:32:33 +00:00
e52a11daa0
Now test pointer result _and_ error code on return from QueryInterface, as some implementations return garbage.
1999-01-16 02:37:22 +00:00
c2935a0853
scc and I decided that the explicit keyword was not necessary for the constructor.
...
It never hurts you to make an nsCOMPtr. MJ
1999-01-15 18:57:31 +00:00
a89c841519
defined NSCAP_NEED_UNUSED_VIRTUAL_IMPLEMENTATIONS for GNU 2.7 under linux. Fixed URL leading to the user manual. Noted that ramiro is working on getting all these special #defines handled by the configuration system instead.
1999-01-05 21:07:57 +00:00
d5aa76aba6
added |dont_QueryInterface| facility, re-added the NS_ADDREF macros, though conditionally
1998-12-21 23:26:39 +00:00
38897ed3d0
NSCAP_FEATURE_DONT_ADDREF is no longer optional, and the symbol has gone away. This facility turns out to be needed. Also catch the case when operator= is called on a dereferenced smart pointer. Fixed some documentation.
1998-12-18 03:45:44 +00:00
e42b720fb2
enabled the dont_AddRef feature, for when functions return an already AddRefed result, e.g. |mSmartPtr = dont_AddRef(GetThing());|
1998-12-17 19:32:00 +00:00
cbfe9f5b15
First Checked In.
1998-12-17 19:12:45 +00:00
scc%mozilla.org
brendan%mozilla.org
dbaron%fas.harvard.edu
cls%seawood.org
mkaply%us.ibm.com
waterson%netscape.com
scc%netscape.com
dmose%mozilla.org
dougt%netscape.com
dveditz%netscape.com
jband%netscape.com
shaver%netscape.com
mcafee%netscape.com
ramiro%netscape.com
alecf%netscape.com
mjudge%netscape.com