# php id string

- Thread: https://externals.io/message/46224
- Original archive: https://news-web.php.net/php.internals/46224
- Started by jvlad on 2009-11-29T10:29:42+00:00
- Messages: 14
- Reply by email to internals@lists.php.net
- More for agents: search with https://externals.io/search.md?q={query} · RFCs at https://externals.io/rfc.md · JSON API at https://externals.io/api · documentation at https://externals.io/llms.txt

---

## #46224 — jvlad — 2009-11-29T10:29:42+00:00

URL: https://externals.io/message/46224#46224

Hi all,

Starting with version 5.3 php checks id string when it loads the extensions 
to match its own one and it also shows this string in PHP Extension Build 
line of `phpinfo()`. That's great. This line contains api#, threadsafe, and 
compiler. So it's almost all important thigs to check and make sure that a 
particular module is binary-compatible with php core. All things, except 
just one, the CPU. It's known that Windows runs on many CPUs, Solaris runs 
fine under sparc, sparc64, x86, and x86_64. Needless to mention linux and 
*bsd systems (I guess they are running on everything). Why not to add what 
 `phpinfo()` shows in Architecture, to the id string? Are there any reasons not 
to do this?

-jvlad

---

## #46225 — Jess Portnoy — 2009-11-29T11:37:57+00:00

URL: https://externals.io/message/46224#46225 · In reply to: #46224

Hello,

Have you considered cases such as universal MAC/Darwin builds? The 
universal build method [used only by Apple but still, many PHP 
developers do run MAC] means you have several archs bundled together in 
the same binary, and, a binary built 2 ways [i386 and PPC or i386 and 
x86_64 or even 4 ways for that matter] can work on any of these archs. 
How do you suggest to handle that?

May the source be with you,
Best regards,
Jess Portnoy



jvlad wrote:
> Hi all,
>
> Starting with version 5.3 php checks id string when it loads the extensions 
> to match its own one and it also shows this string in PHP Extension Build 
> line of `phpinfo()`. That's great. This line contains api#, threadsafe, and 
> compiler. So it's almost all important thigs to check and make sure that a 
> particular module is binary-compatible with php core. All things, except 
> just one, the CPU. It's known that Windows runs on many CPUs, Solaris runs 
> fine under sparc, sparc64, x86, and x86_64. Needless to mention linux and 
> *bsd systems (I guess they are running on everything). Why not to add what 
> `phpinfo()` shows in Architecture, to the id string? Are there any reasons not 
> to do this?
>
> -jvlad

---

## #46226 — Jess Portnoy — 2009-11-29T11:41:02+00:00

URL: https://externals.io/message/46224#46226 · In reply to: #46225

And, though I'm sure everyone on the list is capable of finding this URL 
on their own, a short explanation:
http://en.wikipedia.org/wiki/Universal_binary

May the source be with you,
Best regards,
Jess Portnoy



Jess Portnoy wrote:
> Hello,
>
> Have you considered cases such as universal MAC/Darwin builds? The 
> universal build method [used only by Apple but still, many PHP 
> developers do run MAC] means you have several archs bundled together 
> in the same binary, and, a binary built 2 ways [i386 and PPC or i386 
> and x86_64 or even 4 ways for that matter] can work on any of these 
> archs. How do you suggest to handle that?
>
> May the source be with you,
> Best regards,
> Jess Portnoy
>
>
>
> jvlad wrote:
> > […]

---

## #46227 — Pierre Joye — 2009-11-29T11:43:45+00:00

URL: https://externals.io/message/46224#46227 · In reply to: #46224

hi,

This info is available in phpinfo on windows and I would like to add
it in the "php -v" output as well. I'm not sure how we can safely rely
on this info on other platforms but that's definitively something we
should try to do.

Cheers,

> Hi all,
>
> Starting with version 5.3 php checks id string when it loads the extensions
> to match its own one and it also shows this string in PHP Extension Build
> line of `phpinfo()`. That's great. This line contains api#, threadsafe, and
> compiler. So it's almost all important thigs to check and make sure that a
> particular module is binary-compatible with php core. All things, except
> just one, the CPU. It's known that Windows runs on many CPUs, Solaris runs
> fine under sparc, sparc64, x86, and x86_64. Needless to mention linux and
> *bsd systems (I guess they are running on everything). Why not to add what
> `phpinfo()` shows in Architecture, to the id string? Are there any reasons not
> to do this?
>
> -jvlad
>
>
>
> --
> 
> 
>
>



-- 
Pierre

http://blog.thepimp.net | http://www.libgd.org

---

## #46228 — Jess Portnoy — 2009-11-29T12:19:44+00:00

URL: https://externals.io/message/46224#46228 · In reply to: #46227

Perhaps it would be wise to display both the build arch and the current 
arch on which its running?
I used the Darwin/MAC universal build example before but even on Windows 
and *nix as well when you think about it, one can run a 32bit binary on 
a 64bit OS, usually provided the stack below [Apache, etc] is also 32 bit.
So, unlike the PHP_COMPILER_ID check, which makes sense as the various 
VCs are declared as not quite compatible, I think in the case of 
different archs this would be a mistake, just displaying the gathered 
arch info I can see no harm in though...

May the source be with you,
Best regards,
Jess Portnoy



Pierre Joye wrote:
> hi,
>
> This info is available in phpinfo on windows and I would like to add
> it in the "php -v" output as well. I'm not sure how we can safely rely
> on this info on other platforms but that's definitively something we
> should try to do.
>
> Cheers,
> 
>   
> > […]

---

## #46229 — Pierre Joye — 2009-11-29T12:29:08+00:00

URL: https://externals.io/message/46224#46229 · In reply to: #46228

> just displaying the gathered arch info I can see no harm in though...

that's what we do.

Cheers,
-- 
Pierre

http://blog.thepimp.net | http://www.libgd.org

---

## #46230 — jvlad — 2009-11-29T13:40:39+00:00

URL: https://externals.io/message/46224#46230 · In reply to: #46228

"Jess Portnoy" <jess@zend.com> wrote in message 
news:4B1266E0.7010405@zend.com...
> Perhaps it would be wise to display both the build arch and the current 
> arch on which its running?
> I used the Darwin/MAC universal build example before but even on Windows 
> and *nix as well when you think about it, one can run a 32bit binary on a 
> 64bit OS, usually provided the stack below [Apache, etc] is also 32 bit.
> So, unlike the PHP_COMPILER_ID check, which makes sense as the various VCs 
> are declared as not quite compatible, I think in the case of different 
> archs this would be a mistake, just displaying the gathered arch info I 
> can see no harm in though...
>
> May the source be with you,
> Best regards,
> Jess Portnoy
>
>
>
> Pierre Joye wrote:
> > […]

Jess,

Current platform plays no role. If it can't run a particular php build, 
there is nothing to care of.
What I do care of is ABA which depends on the compile-time arch and nothing 
else.
It's my understanding that id-string is a part of the technology to make 
sure that extensions
are compatible with core.

-jvlad

---

## #46231 — jvlad — 2009-11-29T13:44:39+00:00

URL: https://externals.io/message/46224#46231 · In reply to: #46225

"Jess Portnoy" <jess@zend.com> wrote in message 
news:4B125D15.4040308@zend.com...
> Hello,
>
> Have you considered cases such as universal MAC/Darwin builds? The 
> universal build method [used only by Apple but still, many PHP developers 
> do run MAC] means you have several archs bundled together in the same 
> binary, and, a binary built 2 ways [i386 and PPC or i386 and x86_64 or 
> even 4 ways for that matter] can work on any of these archs. How do you 
> suggest to handle that?
>
> May the source be with you,
> Best regards,
> Jess Portnoy
>
>
>
> jvlad wrote:
> > […]

Jess,
I'm not sure that people who compile php under MAC will do unitversal 
binaries.
At least by-default gcc does not produce them and seems an extra tool is 
required.
Finally, if you have 4 binaries in one file, why don't you have 4 id-strings 
for them?

-jvlad

---

## #46232 — Jess Portnoy — 2009-11-29T13:49:55+00:00

URL: https://externals.io/message/46224#46232 · In reply to: #46230

Apple ships their MAC OS with GCC that is capable of building universal 
binaries.
Most MAC users expect packages to be built universal. Trust me, I also 
hate it but its true...
If what you want is just to ensure the extensions are built for the same 
architecture as the PHP core, this I can understand, I'm just saying you 
need to take multiple archs bundled together under consideration. Also, 
I don't know if deciding for the dynamic loader if something can be 
loaded is so wise, if it can great, if not, it will yell at you anyhow..

May the source be with you,
Best regards,
Jess Portnoy



jvlad wrote:
> "Jess Portnoy" <jess@zend.com> wrote in message 
> news:4B1266E0.7010405@zend.com...
>   
> > […]
>
> Jess,
>
> Current platform plays no role. If it can't run a particular php build, 
> there is nothing to care of.
> What I do care of is ABA which depends on the compile-time arch and nothing 
> else.
> It's my understanding that id-string is a part of the technology to make 
> sure that extensions
> are compatible with core.
>
> -jvlad

---

## #46233 — Jess Portnoy — 2009-11-29T13:51:31+00:00

URL: https://externals.io/message/46224#46233 · In reply to: #46232

BTW, Macports also support building PHP in universal mode.

May the source be with you,
Best regards,
Jess Portnoy



Jess Portnoy wrote:
> Apple ships their MAC OS with GCC that is capable of building 
> universal binaries.
> Most MAC users expect packages to be built universal. Trust me, I also 
> hate it but its true...
> If what you want is just to ensure the extensions are built for the 
> same architecture as the PHP core, this I can understand, I'm just 
> saying you need to take multiple archs bundled together under 
> consideration. Also, I don't know if deciding for the dynamic loader 
> if something can be loaded is so wise, if it can great, if not, it 
> will yell at you anyhow..
>
> May the source be with you,
> Best regards,
> Jess Portnoy
>
>
>
> jvlad wrote:
> > […]

---

## #46234 — jvlad — 2009-11-29T13:57:41+00:00

URL: https://externals.io/message/46224#46234 · In reply to: #46229

>
> > […]
>
> that's what we do.
>
hmm, 32bit php binary v5.3.1 under W7-64bit displays
Windows NT MyHostName 6.1 build 7600 ((null)) i586

If I needed to know current system, I'd prefer to see
SYSTEMNAME,VERSION,HOST,CPU
for example:
Windows Seven,6.1.0.7600,MyHostName,x86_64

and btw, it should be clear that Architecture is compile-time thing,
not run-time, so there is one more reason to move it to id-string.



-jvlad

---

## #46235 — jvlad — 2009-11-29T14:23:26+00:00

URL: https://externals.io/message/46224#46235 · In reply to: #46232

"Jess Portnoy" <jess@zend.com> wrote in message 
news:4B127C03.1080707@zend.com...
> Apple ships their MAC OS with GCC that is capable of building universal 
> binaries.
> Most MAC users expect packages to be built universal. Trust me, I also 
> hate it but its true...
> If what you want is just to ensure the extensions are built for the same 
> architecture as the PHP core, this I can understand, I'm just saying you 
> need to take multiple archs bundled together under consideration. Also, I 
> don't know if deciding for the dynamic loader if something can be loaded 
> is so wise, if it can great, if not, it will yell at you anyhow..
>
> May the source be with you,
> Best regards,
> Jess Portnoy

Jess,
Honestly, I see no problem with universal binaries. After all
they are not that "universal" at all.
Perhaps you missed another reply where I asked - if you have
4 binaries in one file, why don't you have 4 id strings for them?

and yes, many companies that provide their php extensions in pre-compiled 
form like Zend, ionCube, and
many others are interested in a clear way on how to detect whether installed 
php is supported and
what module among hunderd others to pick and install. So, you usee it's not 
a way to dictate
the loader what to load. It's  a way for installer to be wise enough.

-jvlad

---

## #46236 — Jess Portnoy — 2009-11-29T14:43:11+00:00

URL: https://externals.io/message/46224#46236 · In reply to: #46235

Hi,

Please see below.

May the source be with you,
Best regards,
Jess Portnoy



jvlad wrote:
> "Jess Portnoy" <jess@zend.com> wrote in message 
> news:4B127C03.1080707@zend.com...
>   
> > […]
>
> Jess,
> Honestly, I see no problem with universal binaries. After all
> they are not that "universal" at all.
> Perhaps you missed another reply where I asked - if you have
> 4 binaries in one file, why don't you have 4 id strings for them?
>   
The problem I see [and it can certainly be that we just don't understand 
each other's intentions :)], is this:
If the PHP Core [the Apache module, CLI/CGI binary, etc] is built for 
instance as a 4 way universal binary [32bit and 64bit, Intel and PPC], 
the machine itself is, lets say, a 64bit Intel machine and the extension 
being examined is built 2 ways, say, 32bit and 64bit, how do you suggest 
we compare them?
I mean, the extension should be able to load fine even though its just 2 
way because we are NOT running PPC here..
> and yes, many companies that provide their php extensions in pre-compiled 
> form like Zend, ionCube, and
> many others are interested in a clear way on how to detect whether installed 
> php is supported and
> what module among hunderd others to pick and install. So, you usee it's not 
> a way to dictate
> the loader what to load. It's  a way for installer to be wise enough.
>   
The installer really has nothing to do here the way I see it, a person 
may later take an extension built by some other party, this extension 
which technically could load just fine, will be rejected due the Engine 
test, as is the situation today when compiling the extension with a VC 
version different than core was compiled with. If you're just talking 
about warning [not the case these days with the compiler ID check, then 
perhaps].
> -jvlad

---

## #46237 — jvlad — 2009-11-29T15:46:14+00:00

URL: https://externals.io/message/46224#46237 · In reply to: #46236

> > […]
> The problem I see [and it can certainly be that we just don't understand 
> each other's intentions :)], is this:
> If the PHP Core [the Apache module, CLI/CGI binary, etc] is built for 
> instance as a 4 way universal binary [32bit and 64bit, Intel and PPC], the 
> machine itself is, lets say, a 64bit Intel machine and the extension being 
> examined is built 2 ways, say, 32bit and 64bit, how do you suggest we 
> compare them?
> I mean, the extension should be able to load fine even though its just 2 
> way because we are NOT running PPC here..

Well, for your universal extension you have 4 ids. If you care of 
installers, they
should know all 4 ids for the module and pick this module if any one of them
matches. If you care of php core, it will see only the loaded id after the 
loader has
made its decision and picked right part from the hamburger.

> > […]
> The installer really has nothing to do here the way I see it, a person may 
> later take an extension built by some other party, this extension which 
> technically could load just fine, will be rejected due the Engine test, as 
> is the situation today when compiling the extension with a VC version 
> different than core was compiled with.

In fact, the IDs will be different in this case - vc6 and vc9 for visual 
studio 6 and
2008 respectively. So, still no problems. ID clearly indicates whether an 
extension
can be used with particular php core distribution. It's only necessary to 
add CPU :)

-jvlad
