Re: question about linker

On 14/12/2024 04:36, Waldek Hebisch wrote:

David Brown <david.brown@hesbynett.no> wrote:

On 13/12/2024 15:20, Waldek Hebisch wrote:

Janis Papanagnou <janis_papanagnou+ng@hotmail.com> wrote:

>
What I did care about was; about whom Waldek spoke when formulating
"explicit endorsement from gawk developer" - I asked "Who was that?"
>
Because I was surprised by his statement and curious where he got
that idea from. Since the statement I found gave a fairly different
picture. YMMV. - And since I know Arnold - the head of the GNU Awk
maintainers - from various public and private conversations, Waldek's
interpretation (and yours, of course) irritated me, to say the least.

>
You can ask Arnold what he meant.  I saw reasonably recent post
by him implying the he is still using tcc.  And a seqence of
posts from 2013, where he reported problem with tcc and later
wrote that new version (containing fixes) works to compile
gawk.  His messages indicated that he cared about compile
speed and considerd tcc to be fast.
>
  From message about to tinycc-devel dated 'Sun, 06 Jan 2013':
>
: It is quite fast, which is a significant pleasure compared to
: gcc or clang.
>
There he reports problems, later message confirms that changes
to tcc fixed the them.
>
Let me summarize facts as I see them:
- he used tcc to develop gawk

>
I don't see that from what you have written here.  Perhaps it is true,
but unless I have missed something, you haven't given evidence of that.
Note that "used tcc /while/ developing gawk" is not at all the same
thing as "used tcc /to/ develop gawk".  The tools you use to develop
something are your main tools that help you produce good, working code.
gcc is the tool he uses for get correct code - tcc is merely for quick
turnaround testing.

 I do not get what distinction you want to make.  If somebody listens
to music while coding, then "while" looks appropriate.  I develop
a project intending to support compilation by compiler A, B, C, D, E
(that is 5 different compilers), most developement happens using
compiler A, but there are pieces of code specific to other compilers,
which where developed using appropriate compiler.  From time to time
I run test to verify that using each compiler produces expected
results.  It is fair to say that I do small part of developement
using other compilers, but I would say that I use them _to_ develop
the project.  It does not matter that they are not my main compiler.
 Concerning statements about tcc, blurb that Janis quoted literally
says 'using tcc for developement'.  And my reading of the whole
blurb is that he was 'using tcc for developement' (there is more
statements on the net, but I do not think it makes sense for me
to dig them, this should be enough).  I am not a native English
speaker, but for me 'using X for developement of Y' and 'using X
to develop Y' are synonyms, with second being shorter than the
first, while in first skipping Y is more natural.
 

(Again, let me stress that it is possible that he
did use tcc as a main compiler during development, but I don't think you
have shown that.)
>

- he said that tcc can be used to compile gawk

>
Yes.
>

- he complained about speed of gcc/clang and noted that tcc is fast.

>
He said that tcc is "quite fast", faster than gcc or clang, and that he
liked that it was fast.  That is not the same thing as a direct
complaint about the speed of gcc or clang

 Janis snipped that part containing the complaint.  Janis claimed that
complaint does not mean that he liked speed of 'tcc', that is why
I specifically quoted part where he said the he liked the speed.
 

- though clearly he would have
been happier if those compilers had been faster.  (And we all would be
happier if they were faster - even those of us who find gcc fast enough
for our needs.)
>

If that is not an endorsement, than what is?
>

>
It is saying that tcc is a tool you can use to compile gawk, and praise
of its speed relative to gcc and clang.  An endorsement would be saying
that it is the compiler he likes to use or recommends using.
>
If I write a program and say it can run on Linux or Windows, that is not
an endorsement for Windows.

 I would take is as an endorsement for "running the program on Windows".
 Anyway, you wrote:
: No actual developer would care if their code can be compiled by your
: little toy compiler, or even more complete little tools like tcc.
 Gawk developers disprove "no actual developer would care if their
code can be compiled by tcc".  A little grep search discoverd few
tcc-specific lines added to configure machinery.  Google search
discoverd more projects, I did not try to find how many other
project cared and how much, but clearly taken literally
'no actual developer' is false (which you admited in another
message).  It is possible that 'majority of actual developers'
do not care about tcc, I have no hard data, but my feeling is that
developers caring abut tcc are not so small minority.
For me this is quite different from "no actual developer".
Even in non-literal reading IMO "no actual developer"
exagerates smallness of this group.
 

Both JP and DB are clutching at straws.
I really don't know why they so against the concept of a fast, streamlined compiler that cuts some corners. It's not as though there were some decree where you must choose either compiler A or B; you can use both!
Clearly some people (obviously not those two) think the trade-offs are worthwhile. And they know how to mix-and-match A and B to be most productive.
Personally I've also found some problems with TCC in building some programs, eg. missing bits in headers, or sometimes some mysterious error.
I run TCC in order to compare compilation speed, and to that end I will sometimes tweak things (eg. I used my version of windows.h, or compiled preprocessed C in order to avoid a mire of conditional blocks which try and which determine which compile is running, and get it wrong).
Once a program has compiled, I don't remember bugs in TCC's code; it's just a bit slower.
(BTW when I tried to build TCC from source using gcc-O3, I found the resulting compiler even faster than the pre-build binaries! I found that a bit scary so I test only with binaries as supplied.)