148723 - Bad cursor position in all text fields.

Status: RESOLVED DUPLICATE of bug 145267

(Core :: XUL, defect)

Description

[Todd Stock]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.0) Gecko/20020602
BuildID:    Built from source

The cursor appears at the begining of the line.  You can place the real cursor
using the mouse however the blinking line does not follow your typing nor does
it get placed where you mouse press.  Overwrites and what not also seem to be a
bit wierd.  The problem started in RC3 and exists in 1.0 release.

I'm on Linux custom built system using GCC 3.1 GlibC 2.2.5 Kernel 2.4.19 pre17


Reproducible: Always
Steps to Reproduce:
1. Use any text field.
2.
3.

Actual Results:  Wierd positioning of floating blinking cursor.

Expected Results:  They should follow your typing.

Might be specific to my machine being a custom built, but it did not start till RC3.

Comment 1

[R.K.Aa.]

WFM, current trunk CVS, Linux.

Comment 2

[John Levon]

This error is typical of installing a new Mozilla over
an existing installation. You're not doing that are you
Todd ? Does a new profile help ?

What gcc version did you compile RC3 with ?

Comment 3

[Todd Stock]

I am using release tag: MOZILLA_1_0_RELEASE and prior to that
MOZILLA_1_0_RC3_RELEASE.

I do not install it over a prior installation.  After the build is complete I
will  change to the xpinstall/packager directory and run make.  This creates a
gzip'd tar ball. I extract the contents of this into an empty directory.  The
only thing that doesn't change is my .mozilla directory because I have all my
mail and preferences in it.

I built moz RC2 with gcc 3.0.4, moz RC3 with gcc 3.0.4 and later gcc 3.1, and
then moz 1.0 with gcc 3.1.  **I have not tried to build moz RC2 with gcc 3.0.4.**

Comment 4

[Todd Stock]

I just tried the moz 1.0 release with a new/empty profile and the problem still
occurs.

I can try building RC2 with GCC 3.1 and report my findings.  I can also try
building 1.0 with GCC 3.0.4, but that will take a bit longer since I have to
re-create the binaries for GCC 3.0.4.  Please let me know if you need me to try
this...

Comment 5

[Todd Stock]

I did the following tests:
1) I installed the precompile binares for RC3.  I was able to run them because I
have libstdc++ for gcc 2.95 installed on my box as well.  The cursor problem
does not exist.

2) I pulled out RC2 from the CVS trees and compiled them with GCC 3.1.  The
cursor problem exists.

So I can conclude the problem is probably related to GCC 3.1 and might also
exist in GCC 3.0.4 or earlier.  I suppose this bug can be closed.

Comment 6

[John Levon]

Well, we'd rather find out where the problem lies ...
especially if it means we have to report a bug to
the gcc team

Can you try a gcc 3.1 build without using any optimisation ?

Comment 7

[Todd Stock]

Building RC2 and full release 1.0 with the following command line options (debug
enabled) and gcc 3.1 produces a binary set that does not exhibit the cursor
problems.

./configure --prefix=/usr --with-pthreads --with-system-jpeg=/usr
--with-system-zlib=/usr  --with-system-png=/usr  --with-system-mng=/usr
--with-ft-prefix=/usr/X11R6 --enable-crypto --enable-debug --disable-ldap 
--enable-java-supplement

Please let me know if I can help narrow the problem any further.

Comment 8

[Brendan McCarthy]

I have come across the same problem with RC3 and 1.0 using
--enable-optimize="-O3 -march=i686". Building with no optimization did not
result in the bug. I am going to try and see what level of optimization causes
the bug.

I have gcc 3.1 and glibc 2.2.5.

Comment 9

[J. Scott Berg]

Similar configuration, gcc-3.1, glibc-2.2.5, kernel 2.4.18.
Even -O with all relevant -fno-... flags gives the bug.
It only works if -O is not given.
Switching compiler to 2.95.3 solves problem also.
Anyone want to point me to the relevant section of the code?

Comment 10

[R.K.Aa.]

dup of bug 145267?

Comment 11

[Kenneth Herron]

I'm going to resolve this as a dupe of bug 145267. Todd, if you're still having
this problem when you compile with something other than gcc 3.1, please feel
free to reopen.

*** This bug has been marked as a duplicate of 145267 ***