[dpdk-dev] [PATCH v6 9/9] build: generate version.map file for MingW on Windows

[Tal Shnaiderman]

> Subject: Re: [PATCH v6 9/9] build: generate version.map file for MingW on
> Windows
> 
> On Fri, 19 Jun 2020 00:15:46 +0300
> talshn at mellanox.com wrote:
> 
> > From: Tal Shnaiderman <talshn at mellanox.com>
> >
> > The MingW build for Windows has special cases where exported function
> > contain additional prefix:
> >
> > __emutls_v.per_lcore__*
> >
> > To avoid adding those prefixed functions to the version.map file the
> > map_to_def.py script was modified to create a map file for Mingw with
> > the needed changed.
> 
> Thanks for working on this. Your solution is even simpler than we discussed,
> see some comments inline.
> 
> >
> > The file name was changed to map_to_win.py
> >
> > Signed-off-by: Tal Shnaiderman <talshn at mellanox.com>
> > ---
> >  buildtools/{map_to_def.py => map_to_win.py} | 21 ++++++++++++++++-
> ----
> >  buildtools/meson.build                      |  4 ++--
> >  drivers/meson.build                         | 12 +++++++++---
> >  lib/meson.build                             | 15 ++++++++++++---
> >  4 files changed, 39 insertions(+), 13 deletions(-)  rename
> > buildtools/{map_to_def.py => map_to_win.py} (51%)
> >
> > diff --git a/buildtools/map_to_def.py b/buildtools/map_to_win.py
> > similarity index 51% rename from buildtools/map_to_def.py rename to
> > buildtools/map_to_win.py index 6775b54a9d..dfb0748159 100644
> > --- a/buildtools/map_to_def.py
> > +++ b/buildtools/map_to_win.py
> > @@ -13,23 +13,34 @@ def is_function_line(ln):
> >
> >  def main(args):
> >      if not args[1].endswith('version.map') or \
> > -            not args[2].endswith('exports.def'):
> > +            not args[2].endswith('exports.def') and \
> > +            not args[2].endswith('mingw.map'):
> >          return 1
> >
> >  # special case, allow override if an def file already exists
> > alongside map file
> > +# for mingw also replace per_lcore__* to __emutls_v.per_lcore__*
> >      override_file = join(dirname(args[1]), basename(args[2]))
> >      if exists(override_file):
> >          with open(override_file) as f_in:
> > -            functions = f_in.readlines()
> > +            lines = f_in.readlines()
> > +            if args[2].endswith('mingw.map'):
> > +                lines = [l.replace('per_lcore__', '__emutls_v.per_lcore__') for l in
> lines]
> > +            functions = lines
> 
> MinGW GCC linker doesn't fail on undefined symbols, so .map overrides will
> never be needed (and we're going to get rid of overrides soon anyway).
> 

Right, this special case is irrelevant to MinGW, will remove in v7.

> >
> >  # generate def file from map file.
> > -# This works taking indented lines only which end with a ";" and
> > which don't
> > +# For clang this works taking indented lines only which end with a
> > +";" and which don't
> >  # have a colon in them, i.e. the lines defining functions only.
> > +# mingw keeps the original .map file but replaces per_lcore__* to
> > +__emutls_v.per_lcore__*
> >      else:
> >          with open(args[1]) as f_in:
> > -            functions = [ln[:-2] + '\n' for ln in sorted(f_in.readlines())
> > +            lines = f_in.readlines()
> > +            if args[2].endswith('mingw.map'):
> > +                lines = [l.replace('per_lcore__', '__emutls_v.per_lcore__') for l in
> lines]
> > +                functions = lines
> > +            else:
> > +                functions = [ln[:-2] + '\n' for ln in sorted(lines)
> >                           if is_function_line(ln)]
> > -            functions = ["EXPORTS\n"] + functions
> > +                functions = ["EXPORTS\n"] + functions
> 
> Considering the comment above, entire logic for mingw.map is the following:
> 
> if args[2].endswith('mingw.map'):
>     with open(args[1]) as f_in, open(args[2], 'w') as f_out:
> 	f_out.writelines([...replace... for line in f_in.readlines()])
>     return 0
> 
> Wouldn't it be cleaner to move it to the beginning of function or even to a
> separate function dispatched by filename (or even script)? Motivation: soon
> .def logic will become more complicated and Windows-dependent (to
> analyze dumpbin.exe output), intertwining it with .map saves nothing and
> will impede.
> 

IMO Having a separated function for minGW is sufficient, I'll add the change to v7.
Thanks.

> --
> Dmitry Kozlyuk