[R] Followup [Can't install or upgrade the "PKI" package on a Debian testing system]

[Emmanuel Charpentier]

Dear list;

since I've posted the message below, Simon Urbanek has fixed the problem 
(in less than 5 hours, phewww...). The version available on Rforge *is* 
installable with OpenSSL 1.1

Note that you'll have to install it with 
"install.packages('PKI',,'https://www.rforge.net/')", NOT 
install.packages('PKI',,'http://www.rforge.net/'), which, for some 
inscrutable reason (runaway proxying ?) fails the same way as before.

A big Kudos and Thank You to Simon Urbanek, on behalf of not-so-stable 
distributions users !

					Emmanuel Charpentier

Original message follows :
------------------------------------------------------------------------

Dear list,

It seems that recent changes in openssl somehow broke the installation 
or update of the PKI package. This is probably specific of my setup(s) 
(Debian testing, updated frequently).

Copy of a mail to Simon Urbanek (PKI maintainer), sent 5 days ago 
without reply nor acknowledgement so far :
========================================================================

It seems that recent changes in openssl rendered the PKI package 
uninstallable :

--------------------------------------------------------------------------

> install.packages("PKI")
essai de l'URL 'http://cran.univ-paris1.fr/src/contrib/PKI_0.1-3.tar.gz'
Content type 'application/x-gzip' length 31058 bytes (30 KB)
==================================================
downloaded 30 KB

* installing *source* package ‘PKI’ ...
** package ‘PKI’ correctement décompressé et sommes MD5 vérifiées
** libs
gcc -I/usr/local/sage-7/local/lib/R//include -DNDEBUG      -fpic  -g -O2 
   -c asn1.c -o asn1.o
gcc -I/usr/local/sage-7/local/lib/R//include -DNDEBUG      -fpic  -g -O2 
   -c init.c -o init.o
gcc -I/usr/local/sage-7/local/lib/R//include -DNDEBUG      -fpic  -g -O2 
   -c pki-x509.c -o pki-x509.o
pki-x509.c: In function ‘PKI_extract_key’:
pki-x509.c:136:26: error: dereferencing pointer to incomplete type 
‘EVP_PKEY {aka struct evp_pkey_st}’
      if (EVP_PKEY_type(key->type) != EVP_PKEY_RSA)
                           ^~
pki-x509.c: In function ‘get_cipher’:
pki-x509.c:244:40: error: dereferencing pointer to incomplete type 
‘EVP_CIPHER_CTX {aka struct evp_cipher_ctx_st}’
   ctx = (EVP_CIPHER_CTX*) malloc(sizeof(*ctx));
                                         ^~~~
pki-x509.c: In function ‘PKI_RSAkeygen’:
pki-x509.c:550:5: warning: ‘RSA_generate_key’ is deprecated 
[-Wdeprecated-declarations]
      rsa = RSA_generate_key(bits, 65537, 0, 0);
      ^~~
In file included from /usr/include/openssl/rsa.h:13:0,
                  from pki.h:13,
                  from pki-x509.c:1:
/usr/include/openssl/rsa.h:193:1: note: declared here
  DEPRECATEDIN_0_9_8(RSA *RSA_generate_key(int bits, unsigned long e, void
  ^
/usr/local/sage-7/local/lib/R//etc/Makeconf:132 : la recette pour la 
cible « pki-x509.o » a échouée
make: *** [pki-x509.o] Erreur 1
ERROR: compilation failed for package ‘PKI’
* removing ‘/usr/local/sage-7/local/lib/R/library/PKI’

Les packages source téléchargés sont dans
     ‘/tmp/Rtmpnmt97E/downloaded_packages’
Warning message:
In install.packages("PKI") :
   l'installation du package ‘PKI’ a eu un statut de sortie non nul

--------------------------------------------------------------------------

This problem blocks the installation of rstanarm, brms, rsconnect and 
shinystan among others. Not exactly trivial.

As far as I know, my libssl libraries are all at 1.1.0c, as well as openssl.
========================================================================

New data point : it turns out that this problem also impedes the update 
of PKI : my running R installatin still has PKI 1.3, which seems enough 
for rstanarm, brms, rsconnect and shinystan to run and install/upgrade.

However, on a new installation of R (installed in Sage), PKI can't be 
installed. I tried to install PKI 1.3 fro Rforge, with no success.

- Did someone already had this problem ?
- Has he/she been able to work around it ? If so, How ?
- Is my mail to Simon Urbanek sufficient as a bug report ? If not, what
    should I do ?

Sincerely yours,

					Emanuel Charpentier

PS : If possible, I'd appreciate to be CC'd of your answers : I'm not on 
the list and have had trouble subscribing "reasonably". I'm following it 
through the mail archives.