Re: [ft-devel] success?

[Parth Wazurkar]

>
> >> But, even after setting all the metric values and bitmap, it is not
> >> displaying a glyph in `ftview' and just displaying a blank output.
>
> Thanks to your new branch in `freetype2-demos' I can finally test the
> GF+TFM combo...
>
> Assuming that you are running a GNU/Linux box: Have you ever used
> `valgrind', as I've recommended to you a few times?
>

Yes I have :-), Also if you see the `cleaned' branch and debug it with
`valgrind'
you will find it completely error free. These errors by `valgrind' are
reported
after I changed the logic to load bitmaps on demand, and as reported by you
because of the double free in `tfm module'. Sorry for that, I'll fix it :-)

  FT2_DEBUG=any:7 valgrind ./ftview 20 cmr10.600gf &> cmr10.600gf.log
>
> Doing this you can immediately see in `cmr10.600gf.log' that there is
> a double free:
>
>   ...
>   TFM_Read_Metrics: TFM Metric Information:
> Check Sum  : 1274110073
> Design Size: 10
> Begin Char : 0
> End Char   : 127
> font_bbx_w : 10
> font_bbx_h : 10
> slant  : 0
>   Invalid read of size 8
>  at 0x412F6D: FT_Stream_Free (ftobjs.c:251)
>  by 0x41718E: FT_Attach_Stream (ftobjs.c:2711)
>  by 0x417089: FT_Attach_File (ftobjs.c:2672)
>  by 0x40A756: my_face_requester (ftcommon.c:267)
>  by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
>  by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
>  by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
>  by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
>  by 0x406FC4: event_font_change (ftview.c:1039)
>  by 0x408C10: main (ftview.c:1904)
>Address 0x79a80c8 is 56 bytes inside a block of size 80 free'd
>  at 0x4C2B35C: free (in /usr/lib64/valgrind/vgpreload_
> memcheck-amd64-linux.so)
>  by 0x4DF876: ft_free (ftsystem.c:138)
>  by 0x42256C: ft_mem_free (ftutil.c:174)
>  by 0x4DDE84: tfm_close (tfmobjs.c:136)
>  by 0x47F8F1: TFM_Read_Metrics (gfdrivr.c:511)
>  by 0x417154: FT_Attach_Stream (ftobjs.c:2708)
>  by 0x417089: FT_Attach_File (ftobjs.c:2672)
>  by 0x40A756: my_face_requester (ftcommon.c:267)
>  by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
>  by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
>  by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
>  by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
>Block was alloc'd at
>  at 0x4C2A110: malloc (in /usr/lib64/valgrind/vgpreload_
> memcheck-amd64-linux.so)
>  by 0x4DF82B: ft_alloc (ftsystem.c:76)
>  by 0x42229F: ft_mem_qalloc (ftutil.c:76)
>  by 0x40: ft_mem_alloc (ftutil.c:55)
>  by 0x412E2C: FT_Stream_New (ftobjs.c:197)
>  by 0x41710B: FT_Attach_Stream (ftobjs.c:2698)
>  by 0x417089: FT_Attach_File (ftobjs.c:2672)
>  by 0x40A756: my_face_requester (ftcommon.c:267)
>  by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
>  by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
>  by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
>  by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
>
> `FT_Attach_Stream' frees the stream passed as an argument to
> `TFM_Read_Metrics' by itself.  While you have modeled your code after
> `afm_parser_init', you've missed that this function allocates a new
> stream but your code doesn't.  This means that in `tfm_close' you must
> not close the stream since this happens later on.


> Applying this fix
>
>   diff --git a/src/tfm/tfmobjs.c b/src/tfm/tfmobjs.c
>   index 8cd3b5bf5..c03873e7a 100644
>   --- a/src/tfm/tfmobjs.c
>   +++ b/src/tfm/tfmobjs.c
>   @@ -131,9 +131,10 @@
>  FT_LOCAL( void )
>  tfm_close( TFM_Parser  parser )
>  {
>   -FT_Memory  memory = parser->memory;
>   +FT_UNUSED( parser );
>
>   -FT_FREE( parser->stream );
>   +
>   +/* nothing */
>  }
>
> helps.
>

Done!


>
> Note that there are still memory leaks in your driver.
>
>   FT2_DEBUG=any:7 \
>   valgrind --leak-check=full \
>--show-leak-kinds=all \
>./ftview 20 cmr10.600gf
>
>   80 bytes in 1 blocks are definitely lost in loss record 42 of 73
>  at 0x4C2C240: calloc (in /usr/lib64/valgrind/vgpreload_
> memcheck-amd64-linux.so)
>  by 0x4DE498: tfm_parse_metrics (tfmobjs.c:282)
>  by 0x47F85E: TFM_Read_Metrics (gfdrivr.c:497)
>  by 0x417154: FT_Attach_Stream (ftobjs.c:2708)
>  by 0x417089: FT_Attach_File (ftobjs.c:2672)
>  by 0x40A756: my_face_requester (ftcommon.c:267)
>  by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
>  by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
>  by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
>  by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
>  by 0x406FC4: event_font_change (ftview.c:1039)
>  by 0x408C10: main (ftview.c:1904)
>
>   128 bytes in 1 blocks are definitely lost in loss record 44 of 73
>  at 0x4C2C240: calloc (in 

Re: [ft-devel] success?

[Werner LEMBERG]

>   FT2_DEBUG=any:7 valgrind ./ftview 20 cmr10.600gf &> cmr10.600gf.log

For other people reading this I forgot to mention that you need
`cmr10.tfm' in the same directory as `cmr10.600gf' to see the
problems.


Werner

___
Freetype-devel mailing list
Freetype-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/freetype-devel

Re: [ft-devel] success?

[Werner LEMBERG]

>> But, even after setting all the metric values and bitmap, it is not
>> displaying a glyph in `ftview' and just displaying a blank output.

Thanks to your new branch in `freetype2-demos' I can finally test the
GF+TFM combo...

Assuming that you are running a GNU/Linux box: Have you ever used
`valgrind', as I've recommended to you a few times?

  FT2_DEBUG=any:7 valgrind ./ftview 20 cmr10.600gf &> cmr10.600gf.log

Doing this you can immediately see in `cmr10.600gf.log' that there is
a double free:

  ...
  TFM_Read_Metrics: TFM Metric Information:
Check Sum  : 1274110073
Design Size: 10
Begin Char : 0
End Char   : 127
font_bbx_w : 10
font_bbx_h : 10
slant  : 0
  Invalid read of size 8
 at 0x412F6D: FT_Stream_Free (ftobjs.c:251)
 by 0x41718E: FT_Attach_Stream (ftobjs.c:2711)
 by 0x417089: FT_Attach_File (ftobjs.c:2672)
 by 0x40A756: my_face_requester (ftcommon.c:267)
 by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
 by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
 by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
 by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
 by 0x406FC4: event_font_change (ftview.c:1039)
 by 0x408C10: main (ftview.c:1904)
   Address 0x79a80c8 is 56 bytes inside a block of size 80 free'd
 at 0x4C2B35C: free (in 
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
 by 0x4DF876: ft_free (ftsystem.c:138)
 by 0x42256C: ft_mem_free (ftutil.c:174)
 by 0x4DDE84: tfm_close (tfmobjs.c:136)
 by 0x47F8F1: TFM_Read_Metrics (gfdrivr.c:511)
 by 0x417154: FT_Attach_Stream (ftobjs.c:2708)
 by 0x417089: FT_Attach_File (ftobjs.c:2672)
 by 0x40A756: my_face_requester (ftcommon.c:267)
 by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
 by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
 by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
 by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
   Block was alloc'd at
 at 0x4C2A110: malloc (in 
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
 by 0x4DF82B: ft_alloc (ftsystem.c:76)
 by 0x42229F: ft_mem_qalloc (ftutil.c:76)
 by 0x40: ft_mem_alloc (ftutil.c:55)
 by 0x412E2C: FT_Stream_New (ftobjs.c:197)
 by 0x41710B: FT_Attach_Stream (ftobjs.c:2698)
 by 0x417089: FT_Attach_File (ftobjs.c:2672)
 by 0x40A756: my_face_requester (ftcommon.c:267)
 by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
 by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
 by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
 by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)

`FT_Attach_Stream' frees the stream passed as an argument to
`TFM_Read_Metrics' by itself.  While you have modeled your code after
`afm_parser_init', you've missed that this function allocates a new
stream but your code doesn't.  This means that in `tfm_close' you must
not close the stream since this happens later on.

Applying this fix

  diff --git a/src/tfm/tfmobjs.c b/src/tfm/tfmobjs.c
  index 8cd3b5bf5..c03873e7a 100644
  --- a/src/tfm/tfmobjs.c
  +++ b/src/tfm/tfmobjs.c
  @@ -131,9 +131,10 @@
 FT_LOCAL( void )
 tfm_close( TFM_Parser  parser )
 {
  -FT_Memory  memory = parser->memory;
  +FT_UNUSED( parser );
 
  -FT_FREE( parser->stream );
  +
  +/* nothing */
 }

helps.

Note that there are still memory leaks in your driver.

  FT2_DEBUG=any:7 \
  valgrind --leak-check=full \
   --show-leak-kinds=all \
   ./ftview 20 cmr10.600gf

  80 bytes in 1 blocks are definitely lost in loss record 42 of 73
 at 0x4C2C240: calloc (in 
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
 by 0x4DE498: tfm_parse_metrics (tfmobjs.c:282)
 by 0x47F85E: TFM_Read_Metrics (gfdrivr.c:497)
 by 0x417154: FT_Attach_Stream (ftobjs.c:2708)
 by 0x417089: FT_Attach_File (ftobjs.c:2672)
 by 0x40A756: my_face_requester (ftcommon.c:267)
 by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
 by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
 by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
 by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
 by 0x406FC4: event_font_change (ftview.c:1039)
 by 0x408C10: main (ftview.c:1904)

  128 bytes in 1 blocks are definitely lost in loss record 44 of 73
 at 0x4C2C240: calloc (in 
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
 by 0x4DE480: tfm_parse_metrics (tfmobjs.c:281)
 by 0x47F85E: TFM_Read_Metrics (gfdrivr.c:497)
 by 0x417154: FT_Attach_Stream (ftobjs.c:2708)
 by 0x417089: FT_Attach_File (ftobjs.c:2672)
 by 0x40A756: my_face_requester (ftcommon.c:267)
 by 0x4BC018: ftc_face_node_init (ftcmanag.c:243)
 by 0x4BCD50: FTC_MruList_New (ftcmru.c:269)
 by 0x4BC201: FTC_Manager_LookupFace (ftcmanag.c:325)
 by 0x40B0FC: FTDemo_Set_Current_Font (ftcommon.c:551)
 by 0x406FC4: 

Re: [ft-devel] success with branch `kushal-work'

[k kushal]

> On 14-Aug-2017, at 4:30 AM, Werner LEMBERG  wrote:
> 
> 
> Kushal,
> 
> 
> finally, I had success!  Attached is the script I'm using; it should
> run out of the box (note the minor fixes to `make tests’).
Awesome. Will update the script.
> Regarding the HTML presentation, there is still a lot of room for
> visual improvements.  Given that GSoC ends soon, I ask you for new
> commits :-) 
The HTML part is almost complete. I didn’t commit partly completed
Code in this part. 
> In particular, you are generating sprites, but I don't see
> them in action.  Am I missing something?
I have the code. Refining and fixing things. 
> BTW, I suggest that a glyph image appears magnified by a factor 10 or
> so if the cursor hovers over it.  What do you think
Did it. Will update the same :) 

PS: Will update ASAP (in next 36 hrs)
> 
>Werner
> #!/bin/sh
> #
> # This script builds up an environment in the current directory for testing
> # Kushal's font testing stuff.
> #
> # For building, we use the FreeType git repository.  You also need GNU
> # aclocal, autoconf, libtool, and GNU sed.
> 
> 
> # The FreeType git repository URL.
> FREETYPE_REPO=file:///home/wl/git/freetype/freetype2
> 
> 
> #
> # Nothing to configure below.
> #
> 
> # Clean up, just to be sure.
> rm -rf base
> rm -rf test
> 
> 
> # We check out and build version 2.6.5 as the baseline.
> git clone --mirror $FREETYPE_REPO base/.git
> cd base
> git config --unset core.bare
> git config receive.denyCurrentBranch updateInstead
> git checkout -b VER-2-6-5 VER-2-6-5
> 
> # Activate subpixel rendering.
> sed -i "s|/\* *\(#define  *FT_CONFIG_OPTION_SUBPIXEL_RENDERING\) *\*/|\1|" \
> include/freetype/config/ftoption.h
> 
> ./autogen.sh
> ./configure --enable-shared --disable-static
> make
> 
> BASE_DLL=`pwd`/objs/.libs/libfreetype.so
> 
> cd ..
> 
> 
> # We test `master' against the baseline.
> git clone --mirror $FREETYPE_REPO test/.git
> cd test
> git config --unset core.bare
> git config receive.denyCurrentBranch updateInstead
> git checkout -b kushal-work origin/kushal-work
> 
> ./autogen.sh
> ./configure --enable-shared --disable-static
> make
> 
> TEST_DLL=`pwd`/objs/.libs/libfreetype.so
> 
> 
> # Build test program; ensure to not link statically with FreeType.
> cd tests/make_png
> make SRC_LIB= \
> OBJS= \
> LIBS="-lpng -ldl" \
> tests
> 
> 
> # Run tests.
> echo ./tests $BASE_DLL $TEST_DLL test.ttf 20
> ./tests $BASE_DLL $TEST_DLL test.ttf 20
> 
> cd ..
> 
> # eof


___
Freetype-devel mailing list
Freetype-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/freetype-devel