96768 - Need a persistant store for "expensive to calculate" font properties

Status: VERIFIED DUPLICATE of bug 116149

(Core :: Internationalization, defect)

Description

[kill this account]

Bug 96767 proposes to determine the available glyphs in a font. Since this can
be time/CPU expensive it would be good to store the results so they could be
used during later runs without the calculation penalty.

This information does not seem to fit well into preferences so perhaps it 
should have its own file.

Comment 1

[kill this account]

Attachment: patch; new files nsFontCatalogFile.h, nsFontCatalogFile.cpp

Comment 2

[kill this account]

Attachment: patch; sample usage of nsFontCatalogFile (demo only)

Comment 3

[Brendan Eich [:brendan]]

Ah, there this bug is!  Feel free to cc: me any time.  More soon on the patch.

/be

Comment 4

[kill this account]

Attachment: patch; nsNameValuePairDB.{h.cpp} and demo usage in nsFontMetricsGTK

Comment 5

[Brendan Eich [:brendan]]

Shaver wondered whether we might use this db for the XPCOM registry, too. 
Cc'ing him.

/be

Comment 6

[Brendan Eich [:brendan]]

Comment on attachment 47601 [details] [diff] [review]
patch; sample usage of nsFontCatalogFile (demo only)

I dislike using references for out params (non-const refs), because they make the call site's
side effects harder to see than if the caller must use & to pass an address.

Otherwise, looks great -- should I review the main patch?  Sorry I've been swamped by
0.9.5 work.

/be

Comment 7

[kill this account]

okay, I will change to from char*& to char**.

Since the tree will be closed for a few days I'd like to discuss this with
shaver before checking in.

Comment 8

[kill this account]

Attachment: patch; updated GetNextElement() to return needed buffer size

Updated the GetNextElement() method to tell caller how big a buffer
is needed (if current buffer is too small).

GetNextElement() now returns:

  a positive if name and value gotten
  0 if end of file
  a negative number on error
    if (return_val > -NVPDB_MIN_BUFLEN) 
	return_val is an error code
    else 
	return_val is the needed buffer size

Comment 9

[kill this account]

Attachment: sample DB file

Comment 10

[kill this account]

Attachment: patch; additional comments, same code as previous

added comments re: implementing the code to re-get the element if it is
larger than the buffer

Comment 11

[Roy Yokoyama]

Comment on attachment 58359 [details] [diff] [review]
patch; additional comments, same code as previous

/r=yokoyama

Comment 12

[clkao]

I was building with the patch from bug 100570, while compiling
nsNameValuePairDB.o, i got the following error:

 In method `PRBool nsNameValuePairDB::OpenForRead(const char *)':
350: jump to label `OpenForRead_Error'
330:   from here
339:   crosses initialization of `class nsCOMPtr<nsILocalFile> local_file'
332:   crosses initialization of `class nsCAutoString name

simply move the code in OpenForRead_Error into the block calling corresponding
goto solves this.

is my c++ too old? it's gcc 2.95.3

Comment 13

[kill this account]

your compilier is fine.

this version of the code needs attention

I'v been working on it in bug 100570

Comment 14

[kill this account]

Attachment: patch; mozilla/gfx/public dir

Comment 15

[kill this account]

Attachment: patch; mozilla/gfx/src dir

slight improvment to GetNextElement

Comment 16

[kill this account]

I need to check this will compile on Redhat 6.2

Comment 17

[kill this account]

this was implemented in bug 116149

*** This bug has been marked as a duplicate of 116149 ***

Comment 18

[Andreas Becker]

Verifying as dupe, please re-open if someone disagrees.