325 Commits

Author	SHA1	Message	Date
scc%mozilla.org	0b41ff950a	fix self assignment, self append, self replace problems. r=waterson	2000-06-29 22:17:53 +00:00
scc%mozilla.org	e00741f6fd	fix for commercial bustage on Mac (adding |operator const CharT*() const| to |basic_nsLiteralString|). fix for blizzards linux changes needing working |Substring|. r=beard,syd	2000-06-21 05:13:10 +00:00
scc%mozilla.org	76bacb9cc2	fix for bug #43085. r=waterson, a=waterson	2000-06-20 04:25:20 +00:00
scc%mozilla.org	3446bda5c3	fixing OS/2 VisualAge platform bustage, and bad dates in licenses	2000-06-18 05:00:31 +00:00
waterson%netscape.com	78239d38c2	Bug 42836. Handle null string in NS_ConvertUCS2toUTF8::Init().	2000-06-17 01:24:39 +00:00
scc%mozilla.org	9461eb8980	undo the change of the 10th. The code is still in, conditioned on |NS_DONT_USE_CHUNKY_STRING_ALLOCATION|. The performance hit was unacceptable, and checking the original change in on the tip was a mistake. r=rpotts; a=leaf, phil	2000-06-15 22:13:19 +00:00
brendan%mozilla.org	190a0bea06	Fiddle whitespace style.	2000-06-14 04:46:22 +00:00
scc%mozilla.org	2a994e534b	reduce bloat by not chunk-allocating strings (in |nsStr::Alloc|). Clients that don't benefit from this can explicitly call |SetCapacity|. We may re-introduce this functionality in the append of a single character when no space remains, if it turns out to be a problem. This optimization was suggested by profiling done by sfraser, who now promises to find out where it hurts us.	2000-06-12 02:54:22 +00:00
scc%mozilla.org	1d272d98b9	added some #ifdef logic to allow me to turn NEW_STRING_APIS off, so I can continue to profile for bloat; added an implementation specific |operator+()| to resolve ambiguity with promises on some platforms	2000-06-11 04:35:02 +00:00
scc%mozilla.org	84e47f2424	added some #ifdef logic to allow me to turn NEW_STRING_APIS off, so I can continue to profile for bloat; added an implementation specific |operator+()| to resolve ambiguity with promises on some platforms; added a conversion operator (against my best judgement) to |NS_ConvertASCIItoUCS2| to support the construction of literal strings induced by the bad case of the new macro. I could conditionally define this is the opposite direction.	2000-06-11 01:06:14 +00:00
scc%mozilla.org	7f51a498b1	added some #ifdef logic to allow me to NEW_STRING_APIS off, so I can continue to profile for bloat	2000-06-11 01:03:19 +00:00
scc%mozilla.org	85aa1f07df	In an assignment, |SetLength(0); SetLength(newLength);| so that the implementation won't bother to copy any old characters if allocation is required. Fiddled with inlining to reduce bloat ... more experiments needed.	2000-06-11 01:02:12 +00:00
scc%mozilla.org	3adf5c9604	provide real definitions for |NS_LITERAL_STRING| (one of which exploits |wchar_t|); add a range-check assertion to |CharAt|; add another constructor to |nsPromiseConcatenation| to allow per-class |operator+()| (which some compilers need to resolve ambiguity) to work; fiddled with inlining to reduce bloat ... more experiments needed	2000-06-11 01:00:08 +00:00
alecf%netscape.com	a0b04734b4	fix purify errors - uninitialized memory on poorly formatted CID strings, and FMM on nsString::fputs ... r=waterson, scc a=waterson	2000-06-08 23:34:53 +00:00
waterson%netscape.com	5e705488cb	Fix bonehead error warren found: don't re-initialize utf8len; we'll need it to set the string's length!	2000-06-08 23:07:15 +00:00
warren%netscape.com	512c8bf433	Renaming nsIAllocator to nsIMemory (and nsAllocator to nsMemory). API cleanup/freeze. Bug #18433	2000-06-03 09:46:12 +00:00
scc%mozilla.org	752a6ca4ce	Exploiting a user-defined |operator->()| when the result is a pointer to a primitive type is problematic for the OS/2 VisualAge compiler. Eliminate such uses in the new string facilities. These are stragglers missed in an earlier checkin.	2000-06-03 00:38:06 +00:00
scc%mozilla.org	a59a36aebd	VisualAge doesn't like declarations in if/for expressions. r=waterson	2000-06-02 22:35:01 +00:00
waterson%netscape.com	4f43183103	Bugs 41279, 41280, 41290. Implement NS_ConvertUTF8toUCS2, r=ftang. Fix NS_ConvertUCS2toUTF8 to set mLength, r=ftang. Re-implement ToNew[*]String to be less bloaty, r=scc,brendan	2000-06-02 20:06:54 +00:00
scc%mozilla.org	c54f63178e	Exploiting a user-defined |operator->()| when the result is a pointer to a primitive type is problematic for the OS/2 VisualAge compiler. Eliminate such uses in the new string facilities. r=waterson	2000-06-01 06:01:18 +00:00
waterson%netscape.com	0736651092	Bug 39424. Fix Sun Workshop compiler bustage by conditioning NS_READABLE_CAST's funkiness on NEED_CPP_TEMPLATE_CAST_TO_BASE. r=scc	2000-05-27 05:51:56 +00:00
waterson%netscape.com	c82987ae0a	Back out changes that I didn't mean to check in.	2000-05-27 00:17:36 +00:00
waterson%netscape.com	d8892ddb25	Fix boneheaded mistake: I left the return type of NS_ConvertUCS2toUTF8::Init().	2000-05-27 00:17:10 +00:00
waterson%netscape.com	b8f8ba79a7	Bug 40461. Implement NS_ConvertUCS2toUTF8, r=scc,brendan	2000-05-26 22:24:34 +00:00
scc%mozilla.org	00443cdd51	Bug #39963. The NS_LITERAL_[C]STRING macros incorrectly counted the null terminator for a literal string in the length. This may not be the complete solution. r=sdagley	2000-05-21 01:45:19 +00:00
scc%mozilla.org	80e3b6a3ae	Bug #39686. Made an |NS_COUNT|. Put it in the new file "nsAlgorithm.h", to which I also moved |NS_MIN| and |NS_MAX| from "nscore.h", and |copy_string| and |copy_string_backward| from "nsAReadableString.h". Also fixed non-standard include-guards. r={sford, jag}	2000-05-20 20:43:26 +00:00
scc%mozilla.org	dae10b02a8	Adding a new file. Part of the fix for Bug #39686.	2000-05-20 19:57:03 +00:00
scc%mozilla.org	3730f50409	ugh. Fix to placate egcs. |static_cast| should be sufficient for a |void*|, but egcs seems to need |reinterpret_cast|	2000-05-20 19:08:36 +00:00
scc%mozilla.org	82bbf19a82	Fixing platform bustage only caught by SunOS Workshop. Need a |const| in a cast to avoid casting away |const|-ness. r=jdunn	2000-05-20 18:22:05 +00:00
scc%mozilla.org	4814477a24	Checking in patch from Jim Dunn to fix platform bustage on HPUX. In fact, this is the way the code should have been written in the first place due to the C++ standard section 14.6.2(3). Members of the template base class just aren't visible without explicit scoping of some form, typically |this->|. r=scc	2000-05-20 18:19:27 +00:00
scc%netscape.com	fbfcb9d94b	fixed a problem with cut and advancing past the end of a string with an iterator r=waterson	2000-05-16 10:13:23 +00:00
scc%netscape.com	b04f4b0861	turning on the new strings API.	2000-05-16 07:18:04 +00:00
waterson%netscape.com	dbf69aad27	Revert NS_READABLE_CAST. Without the reference, it didn't work.	2000-05-16 07:01:30 +00:00
waterson%netscape.com	5f34d86620	Clean up some of the last warnings.	2000-05-16 00:45:38 +00:00
waterson%netscape.com	fea560d46b	Flip NS_READABLE_CAST back the other way: more important to placate egcs that gcc-2.7.2.3	2000-05-15 23:06:25 +00:00
scc%netscape.com	f7b3e7de41	circular include logic defeated me ... protect the new operator definitions locally	2000-05-15 08:07:25 +00:00
scc%netscape.com	b141c8fd07	oops	2000-05-15 07:37:18 +00:00
scc%netscape.com	626f886483	adding macros and uses to support Solaris' inability to pick operator+	2000-05-15 07:28:00 +00:00
scc%netscape.com	b9cc20ed14	added macro code to help define more comparison operators. Got rid of |typename|. We can live with the warnings, maybe; but windows can't digest it in this context.	2000-05-15 05:14:40 +00:00
scc%netscape.com	a274af8eb7	changes to fix some errors constructing autostrings from null pointers that led to crashes	2000-05-15 03:10:19 +00:00
waterson%netscape.com	58fe0403d8	With NEW_STRING_APIS, turn on definition of string comparison operators to avoid accidental pointer comparisons. r=scc	2000-05-15 01:27:04 +00:00
scc%netscape.com	dc0afa9707	...oops, have to comment out the body too. Sorry.	2000-05-15 00:16:30 +00:00
scc%netscape.com	1280fd6d46	commented out an |NS_WARNING| that will overwhelm you, I'll turn it into an assertion for me only, later. Also got rid of |Equals(*, int)| temporarily, as with |Compare| to catch clients who should be using |EqualsWithConversion|	2000-05-15 00:13:57 +00:00
scc%netscape.com	6965740bc5	disable a not helpful warning for VC++	2000-05-14 23:33:05 +00:00
scc%netscape.com	3e0a608172	comment out |Compare| with a length to help find callers of the old compare function that need to become to |CompareWithConversion|.	2000-05-14 22:32:30 +00:00
waterson%netscape.com	c9c47efc3c	Remove #ifdef that's no longer necessary now that STL headers aren't included.	2000-05-14 00:14:19 +00:00
scc%netscape.com	5bd4efd8d4	waterson and I are giving up on the current build being allowed to even include files related to STL, mostly (but not only) because of Solaris ... by the time someone want to _use_ STL, maybe the compilers will be better, and this fix won't effect them.	2000-05-13 23:59:49 +00:00
scc%netscape.com	37ec015a19	fixed macro-based comparison operators to help with our type-unification problems	2000-05-13 21:39:37 +00:00
scc%netscape.com	018741229b	Added |NS_READABLE_CAST| to clarify what we're doing in places where we face the type-unification error on lame compilers.	2000-05-13 20:29:18 +00:00
scc%netscape.com	e55217832f	getting ready to turn this stuff on, changes related to that: empty strings specified with null pointer, a couple of other fixes.	2000-05-13 20:17:35 +00:00