GitLab

* Bug: In pat, the two "decl" & "desc" parser share their yylval variable,
    and it was declared in both of them, hence the redefinition error.
    Now patched extern in "decl".
* Bug: In scr/main.{h,c} avoid redefinition of NameVdd & NameVss.

* Change: In the AP parser (mbk/alc_pars_p.c), elements are inserted in
    head of list and in the AP driver they are saved in the list order.
    The result is that each time we perform a full read/write cycle, all
    lines in the AP file are reversed. This is annoying to spot differences
    betweeen mofificated files and for git managment as side effect.
      So now reverse all the PHFIG elements list (PHINS, PHCON, PHVIA,
    PHSEG, PHREF) after loading.
      Add functions in mph.h to perform fast chained list reversal.

* Bug: In mbk/src/vel_drive.c, when making the signal names VHDL
    compliants, we should transliterate the SEPAR character, whatever
    it is, by an '_'. It seems that cougar/lynx was not abiding with
    the SEPAR and creates signal names with '.' anyway. So we add a
    test to check for '.' in addition to SEPAR.
      This problem was causing invalid VHDL to be written, with dots
    in signal names.
* Change: In rds/src/etc/cmos.rds, correctly generate GDS for CALUx.

Move extern variables to headers to be compatible with clang

* Bug: In pat/src/pat_decl_y.y the yylval union type for "valu" was int
    instead of long. In pat_desc_y.y it was long, and pat_decl_l.l was
    using the wrong one. That is Flex was writting an int (4 bits) but
    bison was reading a long (8 bits), so the four upper bits were
    random. Dis show only under 64 bits where int and long have
    differnt sizes...
      There is something strange nevertheless in this tool, the flex
    part for pat_desc seems to be missing, but it compiles (and run).

* New: Ported gscr/src placer and channel router for 2 metal only
    technologies (for Hibikino & Make::LSI project).
* Bug: In rfmaccess.c, do not draw rectangle when both size are
    too small instead of only one of them.

* New: In x2y.c, added ability to extract specifically one figure.
    Add vdd/vss terminals for compliance with Alliance.

* Bug: In gds_drive.c, the RDS data structure may contains rectangle of
    null width or height. Those rectangles must not be put into the GDS
    stream as they prevents the GDS parser to re-read it. For now, we
    skip them and issue a warning. Further investigations must be made
    to understand how thoses rectangles gets generateds.
* Bug: In rutpoly.c, some long for managing GDS coordinates where still
    lurking. Now all are replaced by int32_t.

* Bug: In gds_drive.c, the XY GDSII record for storing the instances
    coordinates was using two longs (8 bytes each) instead of int (4 bytes).
    Now use a coord_t to force them to be 32 bits integers.
* Bug: In gds_parse.c, symmetric bug to the above. Read 32 bits integers
    for XY of instances instead of 64 bits integers.