In IM away messages, users can use macros like %n to denote the name. I am passing an away message that is equal to: Hi %n, I'm out to lunch today is %d time is %t The %d and %t seem to always work, but with %n, I get, first time I called it, no replacement was made, and the second time, it replaced the %n twice with the replacement string "SLOGAN621" in this case. Here are some traces in gdb. Breakpoint 2, nsCIMWnd::ProcessAwayMessage (this=0x8cd50a8, msg=0xbfffdcc4, screenName=0x8f05fc0) at /opt/raptor/ns/AIMGlue/src/nsCAimIM.cpp:1137 1137 nsCAutoString msgStr( msg->GetUnicode() ); (gdb) n 1139 nsCOMPtr<nsIDateTimeFormat> gFormat; (gdb) print msgStr $1 = {<nsCString> = {<nsStr> = {mLength = 46, mCapacity = 63, mCharSize = eOneByte, mOwnsBuffer = 0, { mStr = 0xbfffdae4 "Hi %n, I'm out to lunch today is %d time is %t", mUStr = 0xbfffdae4}}, _vptr. = 0x401e6960}, mBuffer = "Hi %n, I'm out to lunch today is %d time is %t\000@ëG\035@\\\000\000\000\200½\215A\220zU\b"} (gdb) n 1140 nsString aResultTime(""); (gdb) n 1141 nsString aResultDate(""); (gdb) n 1149 ) { (gdb) n 1150 msgStr.ReplaceSubstring( "%t", aResultTime.GetUnicode() ); (gdb) n 1151 msgStr.ReplaceSubstring( "%d", aResultDate.GetUnicode() ); (gdb) n 1154 msgStr.ReplaceSubstring( "%n", screenName ); (gdb) print msgStr $2 = {<nsCString> = {<nsStr> = {mLength = 61, mCapacity = 63, mCharSize = eOneByte, mOwnsBuffer = 0, { mStr = 0xbfffdae4 "Hi %n, I'm out to lunch today is 04/20/00 time is 11:30:02 PM", mUStr = 0xbfffdae4}}, _vptr. = 0x401e6960}, mBuffer = "Hi %n, I'm out to lunch today is 04/20/00 time is 11:30:02 PM\000U\b"} (gdb) n 00:08:02 Socket 08CA6660: readable 1156 msg->Assign( msgStr ); (gdb) print msgStr $3 = {<nsCString> = {<nsStr> = {mLength = 61, mCapacity = 63, mCharSize = eOneByte, mOwnsBuffer = 0, { mStr = 0xbfffdae4 "Hi %n, I'm out to lunch today is 04/20/00 time is 11:30:02 PM", mUStr = 0xbfffdae4}}, _vptr. = 0x401e6960}, mBuffer = "Hi %n, I'm out to lunch today is 04/20/00 time is 11:30:02 PM\000U\b"} (gdb) n 1157 return NS_OK; and the case where it replaced twice: Breakpoint 2, nsCIMWnd::ProcessAwayMessage (this=0x8cd50a8, msg=0xbfffdc3c, screenName=0x8ed90d8) at /opt/raptor/ns/AIMGlue/src/nsCAimIM.cpp:1137 1137 nsCAutoString msgStr( msg->GetUnicode() ); (gdb) n 1139 nsCOMPtr<nsIDateTimeFormat> gFormat; (gdb) n 1140 nsString aResultTime(""); (gdb) n 1141 nsString aResultDate(""); (gdb) print msgStr $7 = {<nsCString> = {<nsStr> = {mLength = 46, mCapacity = 63, mCharSize = eOneByte, mOwnsBuffer = 0, { mStr = 0xbfffdb08 "Hi %n, I'm out to lunch today is %d time is %t", mUStr = 0xbfffdb08}}, _vptr. = 0x401e6960}, mBuffer = "Hi %n, I'm out to lunch today is %d time is %t\000¿øQÍ\b.\000\000\000àª\222A\003\000\000"} (gdb) n 1149 ) { (gdb) n 1150 msgStr.ReplaceSubstring( "%t", aResultTime.GetUnicode() ); (gdb) n 1151 msgStr.ReplaceSubstring( "%d", aResultDate.GetUnicode() ); (gdb) n 1154 msgStr.ReplaceSubstring( "%n", screenName ); (gdb) print msgStr $8 = {<nsCString> = {<nsStr> = {mLength = 61, mCapacity = 63, mCharSize = eOneByte, mOwnsBuffer = 0, { mStr = 0xbfffdb08 "Hi %n, I'm out to lunch today is 04/20/00 time is 11:32:15 PM", mUStr = 0xbfffdb08}}, _vptr. = 0x401e6960}, mBuffer = "Hi %n, I'm out to lunch today is 04/20/00 time is 11:32:15 PM\000\000"} (gdb) n 1156 msg->Assign( msgStr ); (gdb) print msgStr $9 = {<nsCString> = {<nsStr> = {mLength = 75, mCapacity = 128, mCharSize = eOneByte, mOwnsBuffer = 1, { mStr = 0x8eca650 "Hi SLOGAN621SLOGAN621, I'm out to lunch today is 04/20/00 time is 11:32:15 ", mUStr = 0x8eca650}}, _vptr. = 0x401e6960}, mBuffer = "Hi %n, I'm out to lunch today is 04/20/00 time is 11:32:15 PM\000\000"}

Looks like the problem happened when we grew the buffer (and capacity) from 63 to 128 and the bits we copied back.

Core functionality needs to work... or else we will sufer all over the place. PDT+

Syd, I think you may be calling this function incorrectly. Both arguments need to already be the same character encoding, and the same encoding as the referent. Additionally, both arguments need to be pointers. Not forcing this explicitly may make you call into the version that takes to |ns[C]Strings| as parameters, which rickg explictly notes in the code //WARNING: This is not working yet!!!!! but no further details of what is wrong or what he intended to happen. Where is the source that exhibits these symptoms?

Yeah, that's the problem. You're just using a routine that is known not to work. Theorectically, the pointer version works. I'd switch to that for now, until I have a reasonable answer for you. Supposedly, rickg is replacing this entire class soon, so I hesitate to fix a routine that will go away, and only one person is calling. Till then, I'm assigning the bug back to you as asked me to.

Actually I'm not quite positive if the patch addresses the problem described in this report. It fixes a bug occuring when replacing e.g. "&" with "&amp;", where the replacement pattern contains the search pattern. This lead to a (more or less) infinite loop...

I guess I need to learn about bugzilla... The attachment is a .tar.gz archive.

Fixed - we've recoded the caller as per scc's suggestion, and this is no longer an issue.

I tested the original error case and it worked fine for me. marking verified as per Jan's request.