On Wed, 2 Mar 2005 02:41:54 +0100, Frank Heckenbach ih8mj@fjf.gnu.de wrote:

James A. Morrison wrote:

...

Hi,

I've posted[1] a diff between December's snapshot and February's snapshot. However, the diff 850K, that seems really large for only 4 changelog entries. Is that the size of the usual diff between patches?

It varies wildly as we don't make releases according to a fixed schedule.

In this case, an additional factor might have been that the bison generated files were no more included in the February release because a bison 2.0 had been released meanwhile (whereas previously we required a somewhat hairy alpha bison version).

Frank

parse.c and rts/autom4te.cache are both large changes, as well as some of the info documentation. I wouldn't be surprised if updating the copyright year in a large number of files also had an effect.

Waldek Hebisch wrote:

If you exclude generated files (note that beside parser and lexer part of documantation is auto generated) the diff is of order 350K.

You may want to use the `-minimal' archives to diff. They still contain a few generated files (temporarily), but not so many, in particular not the info files and (from now on) the bison output.

It is still unusually large, the single biggest component beeing `gpc.c' (actual change to `gpc.c' is quite small, the big size come from merging changes between `gcc.c' from gcc-3.3 and `gcc.c' from `gcc.c').

Of course, I hadn't thought of that.

FYI diff between gpc-20041017 and gpc-20041218 is about 200K and diff between gpc-20040516 and gpc-20041017 is about 370K

By the way Frank, I see the following entries that I have sent you for 20041017 are missing:

ChangeLog:

   * declarations.c, expressions.c, gpc.h, parse.y,
     statements.c, typecheck.c types.c: implemented ISO
     constructors (initializers), added support for packed
     arrays and variant records.

   * objects.c: use 'size_zero_node' (correct type) to initialize
     size field in vmt for abstract objects.

I am not sure if you applied those two:

   * gpc-lex.c, diffs/gcc-3.2.1.diff: workaround to allow use
     of flex 2.5.4

   * predef.c, typecheck.c, types.c: make PG__NEW_STRINGS working,
     use NULL_TREE value of capacity to request undiscriminated
     string schema

I'm not sure I got the PG__NEW_STRINGS patch. Can you send it to me (again)? (AFAIR, I applied some changes of yours WRT PG__NEW_STRINGS, but in the current version, it's still inactive. Or should I just activate it?)

And for 20050217:

   * parse.y, declarations.c, gpc.h, module.c, objects.c, parse.y,
     rts/*.pas, statements.c: handle qualified identifiers
   * config-lang.in, declarations.c, expressions.c, gbe.h, gpc.c, gpc.h,
     gpc-lex.c, gpcpp.c, lang.c, Make-lang.in, options.c, p-tree.def,
     statements.c, typecheck.c, types.c: update to gcc-3.4.x
   * typecheck.c: handle discriminants in schema initilizers
   * types.c: fold constant array references

Sorry, I must have missed the ChangeLog entries. Did you send them by mail, or included in your diff? (Adding them now.)

James A. Morrison wrote:

parse.c and rts/autom4te.cache are both large changes, as well as some of the info documentation. I wouldn't be surprised if updating the copyright year in a large number of files also had an effect.

Right. Around a new year, this also adds.

Frank