I have something like:

---File1.c

// Outside of main
typedef struct {

        int whatever;
        int whatever2;

        } MyType;

---File2.c

extern  MyType;

int some_function()
{
        MyType* MyVar;  // Error from gcc when compiling (not linking
yet)   on this line, "MyVar undeclared"

        return 0;

  MyType athing; /* is an instance of a MyType */

extern MyType athing;  /* and it might get found */

for instance:
typedef struct{ /* <- unnamed struct */
int a;
...

MyStruct var1;
...

The C99 standard states:

 "A typedef declaration does not introduce a new type, only a
  synonym for the type so specified."

No where in section 6.7.7, where typedef is defined, is 'alias' used to
describe what a typedef is. Furthermore, 'alias' and 'synonym' have
different enough meanings such that they don't appear together in the
Thesaurus' I checked.

    Main Entry: 2alias
    Function: noun
    Date: 1605
    : an assumed or additional name

Compare that with the definition of "synonym":

    Date: 15th century
    1: one of two or more words or expressions of the same language
       that have the same or nearly the same meaning in some or all senses
    2: a symbolic or figurative name : METONYM
    3: a taxonomic name rejected as being incorrectly applied or
       incorrect in form -- compare HOMONYM

Given that what typedef does is to create an *additional* name for
some existing type, I submit that the Standard has mis-used the
English word "synonym" here.  Dictionary definitions 2 and 3 do
not seem to apply, and dictionary definition 1 gives "synonym" far
too much latitude -- for instance, if INT_MAX and LONG_MAX are
the same value, and INT_MIN and LONG_MIN are the same value, it
would be correct English to call int and long "synonyms": "two
or more words ... of the same language ... that have ... nearly
the same meaning in some ... senses."

Of course, one can argue that "alias" is pejorative.  Although
m-w.com does not show it above, the word "alias" often implies a
degree of dishonesty.  It also is clearly intended for anthropomorphic
entities[%].  The thesaurus I liked enough to buy ("The Synonym
Finder" by J. I. Rodale, Rodale Press) suggests these for the noun
form of "alias":

    pseudonym, anonym, false name; assumed name, stage name, pen
    name, nom de plum, /Fr. nom de guerre/, allonym; nickname,
    sobriquet.

Yet I would also like to quote the first paragraph of the Introduction
to this book, by Laurence Urdang (28 July 1978):

    Those who work with language know that there is no such thing
    as a true ``synonym.''  Even though the meanings of two words
    may be the same -- or nearly so -- there are three characteristics
    of words that almost never coincide: frequency, distribution,
    and connotation.  /Panthera leo/ and /lion/, /cucurbit/ and
    /squash/, /sodium carbonate/ and /washing soda/ have quite
    different frequencies in English.  We all know that a /house/
    is not a /home/, that not all /women/ are /ladies/, that not
    all /men/ are /gentlemen/; at a more subtle level, we soon
    learn the differences between /motherly/ and /maternal/,
    /fatherly/ and /paternal/, /brotherly/ and /fraternal/.  These
    are connotative differences.

In C programming, however, typedef names are *exact* copies of the
original "true" name of the type.  All differences are syntactic,
even including the fact that qualifiers never "penetrate" a typedef
(so that "typedef char *T; const T b" does not mean the same thing
as "const char *b" but rather "char *const b").  Again, then, the
subtle nuances of "synonym" do not apply to typedef, while those
of "alias" do.  Aside from its slightly wicked connotation (of which
I actually approve: typedefs *are* slightly wicked :-) ), its only
real fault is that it is meant to apply to humans.

Since the word 'alias' and 'synonym' are not interchangeable as
determined by any respected Thesaurus or Dictionary, the potential for
deception is simply to great especially among those who may not be aware
of what is and is not in common, but inaccurate, usage in this
newsgroup.

typedef introduces an alternate name for an existing type. This is
commonly used to avoid typing the word "struct" before struct vars.

My mistake ;-)

    Given that what typedef does is to create an *additional* name for
    some existing type, I submit that the Standard has mis-used the
    English word "synonym" here.

Now, Standards *do* tend to introduce their own jargon, redefining
various words to mean something new.  The C standard's version of
"object" is of course quite different from the dictionary definitions:

    1 a: something material that may be perceived by the senses
    <I see an object in the distance> b : something that when viewed
    stirs a particular emotion (as pity)  <look to the tragic
    loading of this bed... the object poisons sight; let it be hid
    -- Shakespeare>

    2: something mental or physical toward which thought, feeling,
    or action is directed  <an object for study>  <the object of
    my affection>  <delicately carved art objects>

    3 a: the goal or end of an effort or activity : PURPOSE,
    OBJECTIVE <their object is to investigate the matter thoroughly>
    b : a cause for attention or concern  <money is no object>

    4 : a thing that forms an element of or constitutes the subject
    matter of an investigation or science

    5 a : a noun or noun equivalent (as a pronoun, gerund, or
    clause) denoting the goal or result of the action of a verb
    b : a noun or noun equivalent in a prepositional phrase

    synonym see INTENTION

None of these is much like the Standard's definition:

       3.15  Object

       [#1] A region of data storage in the execution  environment,
       the  contents  of  which  can  represent values.  Except for
       bit-fields, objects are composed of contiguous sequences  of
       one  or more bytes, the number, order, and encoding of which
       are either explicitly specified  or  implementation-defined.
       When  referenced,  an  object may be interpreted as having a
       particular type; see 6.2.2.1.

Thus it is not beyond the pale for the C standard to redefine the
English word "synonym" to mean "alias", instead of simply using
the word "alias".  I, however, prefer to use this existing English
word that, in the minds of most of those whom I would consider
sensible, thoughtful, and intelligent readers, will conjure up the
most appropriate image, both via denotation and connotation.  As
such, I will continue to use the word "alias", and encourage others
to do likewise, but also encourage them (and attempt myself) to
note, as needed, that this "alias" is what the Standard calls, in
highly misleading fashion, a "synonym".

If the reader finds that the word "alias" does *not* call forth an
appropriate metaphor, that does not prove that the reader is not
sensible, thoughtful, or intelligent, because after all, everyone
is different.  Still, communication requires some degree of shared
world-view, and effort on both the readers' and writer's parts,
and I believe that on the whole, the writer (singular) should put
forth the most effort in this transaction.  If you find my words
misleading, confusing, or inappropriate, you should feel free to
disagree, whether silently or (as it were) in print: in this
newsgroup, at your choice; and if enough people -- where "enough"
is strictly in my own judgement, of course -- write and say that
my long flowery sentence structure is an impediment to communication,
then I may choose to change.

Dictionaries can be very helpful when you encounter a word in normal use
and don't happen to know what it means, but they make very bad legal
weaponry, especially in a technical discussion.

One should not /rely/ on dictionary definitions for technical terms
(look up 'padding' in your favourite non-technical dictionary if you
don't believe that - 'padding' is used by the Standard, but not defined
therein AFAICT, so it's a valid illustration of my point), since those
terms are very frequently the result of the grafting of new meanings
onto old words, a process which is happening at a much greater rate than
a non-technical dictionary can possibly match. In this case, IMHO,
actual usage is a much better guide.

If they decide to change the wording, then I will accept 'alias' as an
accurate word in relation to typedef.

Until such a time, it is simply inaccurate to use 'alias'. It is
irrelevant that it is in common usage among the regulars. This group
should be as much about the regulars communicating with one another as
it is about them communicating with people who ask questions. To
redefine what a typedef creates a potential deceptive situation, which
is trivially avoided.