1554616 - [patch] add support for lcc compiler (<=1.23)

Status: VERIFIED FIXED

(NSS :: Build, enhancement, P1)

Description

[Michael Shigorin]

Attachment: 0001-freebl-add-lcc-support.patch

NSS currently fails to build with lcc 1.23.12 compiler on 64-bit Elbrus 2000 architecture due to its lack of __int128; not only 3.44 version I've patched but the current hg I've checked the patch to apply against either.

Please find the patch attached.

PS: I hope to revisit this as soon as I can get my hands on lcc 1.24 where __int128 is suppored (not sure when this happens though).

Comment 1

[Michael Shigorin]

Test suite seems to pass for me:

Tests summary:
--------------
Passed:             13297
Failed:             0
Failed with core:   0
ASan failures:      0
Unknown status:     48
TinderboxPrint:Unknown: 48 

Comment 2

[J.C. Jones [:jcj] (he/they)]

Comment on attachment 9067720 [details] [diff] [review]
0001-freebl-add-lcc-support.patch

Review of attachment 9067720 [details] [diff] [review]:
-----------------------------------------------------------------

::: lib/freebl/Makefile
@@ +494,5 @@
>  
>  ifdef USE_64
> +# no __int128 at least up to lcc 1.23 (pretending to be gcc5)
> +# NB: CC_NAME is not defined here
> +ifneq (lcc,$(shell $(CC) -? 2>&1 >/dev/null | sed -e 's/:.*//;1q'))

Everywhere else in the file we put the constant as the second argument to the if. Can you swap it? Then r+ and I'll check it in. Thanks!

Comment 3

[Michael Shigorin]

Attachment: 0001-freebl-add-lcc-support.patch

No problem, thanks for reviewing :-)

Comment 4

[J.C. Jones [:jcj] (he/they)]

Comment on attachment 9070702 [details] [diff] [review]
0001-freebl-add-lcc-support.patch

Review of attachment 9070702 [details] [diff] [review]:
-----------------------------------------------------------------

Thank you!

Comment 5

[J.C. Jones [:jcj] (he/they)]

https://hg.mozilla.org/projects/nss/rev/9162c654d06915f0f15948fbf67d4103a229226f

Comment 6

[Mike Kaganski]

MSVC compiler hangs with cl.exe -? >/dev/null - it waits for user input... apparently redirection of stdout to /dev/null breaks some check there, since redirecting to a dummy file works OK.

Comment 7

[Michael Shigorin]

Mike, that might be worth another ifdef -- and check for CC_NAME in nss/coreconf/Werror.mk as that's where I borrowed the test from (the variable doesn't appear to be defined where I had to inline the test).

Comment 8

[Mike Kaganski]

Well - simply adding stdin to /dev/null also works - so I'd just made something like

ifneq ($(shell $(CC) -? 2>&1 >/dev/null </dev/null | sed -e 's/:.*//;1q'),lcc)

Comment 9

[Michael Shigorin]

Go ahead and post a patch for both occasions as a new bug referencing this one :-)

Comment 10

[Michael Shigorin]

Forgot it: 3.45.0 built on e2k just fine, thank you JC!